Algorithmic processing of digitally-represented images
POPULARITY
Categories
Head on over to https://cell.ver.so/TOE and use coupon code TOE at checkout to save 15% on your first order. Get ready to witness a turning point in mathematical history: in this episode, we dive into the AI breakthroughs that stunned number theorists worldwide. Join us as Professor Yang-Hue Hi discusses the murmuration conjecture, shows how DeepMind, OpenAI, and EpochAI are rewriting the rules of pure math, and reveals what happens when machines start making research-level discoveries faster than any human could. AI is taking us beyond proof straight into the future of discovery. As a listener of TOE you can get a special 20% off discount to The Economist and all it has to offer! Visit https://www.economist.com/toe Join My New Substack (Personal Writings): https://curtjaimungal.substack.com Listen on Spotify: https://open.spotify.com/show/4gL14b92xAErofYQA7bU4e Timestamps: 00:00 Introduction to a New Paradigm 01:34 The Changing Landscape of Research 03:30 Categories of Machine Learning in Mathematics 06:53 Researchers: Birds vs. Hedgehogs 09:36 Personal Experiences with AI in Research 11:44 The Future Role of Academics 14:08 Presentation on the AI Mathematician 16:14 The Role of Intuition in Discovery 18:00 AI's Assistance in Vague Problem Solving 18:48 Newton and AI: A Historical Perspective 20:59 Literature Processing with AI 24:34 Acknowledging Modern Mathematicians 26:54 The Influence of Data on Mathematical Discovery 30:22 The Riemann Hypothesis and Its Implications 31:55 The BST Conjecture and Data Evolution 33:29 Collaborations and AI Limitations 36:04 The Future of Mathematics and AI 38:31 Image Processing and Mathematical Intuition 41:57 Visual Thinking in Mathematics 49:24 AI-Assisted Discovery in Mathematics 51:34 The Murmuration Conjecture and AI Interaction 57:05 Hierarchies of Difficulty 58:43 The Memoration Breakthrough 1:00:28 Understanding the BSD Conjecture 1:01:45 Diophantine Equations Explained 1:03:39 The Cubic Complexity 1:19:03 Neural Networks and Predictions 1:21:36 Breaking the Birch Test 1:24:44 The BSD Conjecture Clarified 1:26:21 The Role of AI in Discovery 1:30:29 The Memoration Phenomenon 1:32:59 PCA Analysis Insights 1:35:50 The Emergence of Memoration 1:38:35 Conjectures and AI's Role 1:41:29 Generalizing Biases in Mathematics 1:44:55 The Future of AI in Mathematics 1:49:28 The Brave New World of Discovery Links Mentioned: - Topology and Physics (book): https://amzn.to/3ZoneEn - Machine Learning in Pure Mathematics and Theoretical Physics (book): https://amzn.to/4k8SXC6 - The Calabi-Yau Landscape (book): https://amzn.to/43DO7H0 - Yang-Hui's bio and published papers: https://www.researchgate.net/profile/Yang-Hui-He - A Triumvirate of AI-Driven Theoretical Discovery (paper): https://arxiv.org/abs/2405.19973 - Edward Frenkel explains the Geometric Langlands Correspondence on TOE: https://www.youtube.com/watch?v=RX1tZv_Nv4Y - Stone Duality (Wiki): https://en.wikipedia.org/wiki/Stone_duality - Summer of Math Exposition: https://some.3b1b.co/ - Machine Learning meets Number Theory: The Data Science of Birch–Swinnerton-Dyer (paper): https://arxiv.org/pdf/1911.02008 - The L-functions and modular forms database: https://www.lmfdb.org/ - Epoch AI FrontierMath: https://epoch.ai/frontiermath/the-benchmark - Mathematical Beauty (article): https://www.quantamagazine.org/mathematical-beauty-truth-and-proof-in-the-age-of-ai-20250430/ SUPPORT: - Become a YouTube Member (Early Access Videos): https://www.youtube.com/channel/UCdWIQh9DGG6uhJk8eyIFl1w/join - Support me on Patreon: https://patreon.com/curtjaimungal - Support me on Crypto: https://commerce.coinbase.com/checkout/de803625-87d3-4300-ab6d-85d4258834a9 - Support me on PayPal: https://www.paypal.com/donate?hosted_button_id=XUBHNMFXUX5S4 SOCIALS: - Twitter: https://twitter.com/TOEwithCurt - Discord Invite: https://discord.com/invite/kBcnfNVwqs #science Learn more about your ad choices. Visit megaphone.fm/adchoices
Как попасть в стартап в Лондоне, писать шейдеры на Metal и не выгореть, работая с видео и графикой? Андрей Михайлов — iOS-разработчик, camera engineer в приложении Lapse и бывший сотрудник Prisma — делится откровенной и насыщенной историей своей карьеры: от первых стартапов в России до релокации в Великобританию.В выпуске:
How can we balance the growing number of satellites on orbit, and the services they provide, with the importance of astronomical science and protecting Dark and Quiet Skies? Reflected light and spectrum interference from satellites have become more problematic for astronomers and Star Gazers alike, but there are mitigation techniques being developed and tested to protect our Night Sky. Understanding the approaches to timely de-orbiting and space traffic awareness play into protecting astronomical instruments and data, and maintaining the beauty of the Night Sky. Join The Aerospace Corporation's Colleen Stover and experts Dr. Lindsay DeMarchi, a “stellar mortician”, and Dr. Samuel Factor, Division of Signal & Image Processing, to discuss the issues and potential solutions around the resilience of our skies. Read more at the American Astronomical Society. This episode is part of the Center for Space Policy and Strategy's series on Resiliency. The series explores various perspectives of what resiliency means from across the space community. The Space Policy Show is produced by The Aerospace Corporation's Center for Space Policy and Strategy. It is a virtual series covering a broad set of topics that span across the space enterprise. CSPS brings together experts from within Aerospace, the government, academia, business, nonprofits, and the national labs. The show and their podcasts are an opportunity to learn about and to stay engaged with the larger space policy community. Subscribe to our YouTube channel to watch all episodes!
In this episode of the Ducks Unlimited Podcast, host Katie Burke and co-host Dr. Mike Brasher are joined by decoy expert Colin McNair of Copley Fine Art Auctions, marking his fifth appearance on the show. Together, they dive into the captivating world of decoy art, focusing on the groundbreaking practice of x-raying decoys to uncover hidden details of craftsmanship. Discover how visual aids and cutting-edge techniques bring new insights to this timeless art form, and hear Colin's expert take on what makes these pieces so unique. Whether you're a seasoned collector, an art enthusiast, or simply curious, this episode is packed with fascinating stories and insider knowledge you won't want to miss.Listen now: www.ducks.org/DUPodcastSend feedback: DUPodcast@ducks.org
In this short podcast, you will learn about the concept of technology in under 60 seconds.
In this episode, we sit down with Donia Chaiehloudj, a Senior Software Engineer at Isovalent, to discuss her diverse career journey, from working in image processing and electronics to becoming a Go developer in the cloud-native space. Donia shares her experiences transitioning into software development, her work with Go and Kubernetes, and her leadership role in the GDG Sophia-Antipolis community. She also touches on her passion for public speaking, open-source contributions, and balancing her career with life as a new mother. This episode is perfect for anyone interested in tech, community building, and career growth in software engineering.00:00 Introduction1:57 What is Donia Doing Today?14:00 Highschool Interests18:34 Engineering School35:31 Internship Work / Software Transition42:15 Metal Health in School50:00 Graduating University / Job Searching58:00 Becoming a Java Developer1:10:00 Public Speaking / Community1:23:50 Contact InformationConnect with Donia: Twitter: https://twitter.com/doniacldLinkedin:https://www.linkedin.com/in/donia-chaiehloudj/Mentioned in today's episode:Isovalent: https://isovalent.com/TinyGo: https://tinygo.org/Want more from Ardan Labs? You can learn Go, Kubernetes, Docker & more through our video training, live events, or through our blog!Online Courses : https://ardanlabs.com/education/ Live Events : https://www.ardanlabs.com/live-training-events/ Blog : https://www.ardanlabs.com/blog Github : https://github.com/ardanlabs
Paola Jaramillo is working at MathWorks as the Technical Manager for the Application Engineering team supporting you and your team of domain experts in the vision creation and implementation of innovative technologies with leading tools and services for professional software development. She is an active participant in high-tech events in Benelux, such as the Smart Systems Industry Summit, Bits & Chips Smart Systems, Dutch Machine Vision Conference, and AutoSens. Her interests are sensor data analytics and autonomous systems, where Signal and Image Processing, Computer Vision, and Deep Learning are commonly used. In her previous experience, she implemented a sensor-based DSP system for Structural Health Monitoring and developed a sensor-based predictive (Machine Learning) system that optimizes energy consumption in office buildings. ONLINE PRESENCE ================
Key Topics & Chapter Markers:Recap from Part 1: The Early Years of AI [00:00:00]AI Architecture & Oracle's Innovation in Hash Joins [00:02:00]Impact of Nature in Creative and Collaborative Work [00:05:00]The Rise of Neural Networks: Language and Image Processing [00:10:00]Sparse and Dense Vectors Explained [00:15:00]Google Translate's Early Approaches & Statistical Methods [00:20:00]TensorFlow vs. PyTorch: Defining the Modern AI Framework [00:30:00]Dot Products, Similarity, and the Concept of Attention [00:35:00]Transformers & The Attention Mechanism Revolution [00:42:00]BERT, GPT, and the Dawn of Transfer Learning [01:00:00]The Road to ChatGPT and OpenAI's Innovations [01:10:00]The Future of AI and Computational Scaling [01:15:00]Share Your Thoughts: Have questions or comments? Drop us a mail at EffortlessPodcastHQ@gmail.com
The Putin regime has exploited the vulnerabilities of global social media platforms, using bots and trolls to promote its narratives. They also use it for coercive manipulation to change public attitudes towards Ukraine aid, and to cause fear, indecision, hesitation and caution among politicians and geo-political commentators. But Russia is not alone in using these methods; once developed the same techniques can be used by other powers, international and local to subvert democratic institutions and to warp public perception of the news and politics in their countries. And in a year of numerous elections, Russian multi-front aggression including info and Cyber poses an enormous threat. ---------- Gundars Bergmanis-Korāts – Principal Scientist at NATO StratCom COE Dr. Gundars Bergmanis-Korats is a Principal Scientist at the NATO Strategic Communications Centre of Excellence (StratCom COE). He holds a PhD in Signal and Image Processing from the University of Lorraine in France. In his role, Gundars concentrates on projects that explore the diverse applications of artificial intelligence (AI) in strategic communications, analysing the risks and opportunities presented by this technology from multiple perspectives, including those of defenders and malicious actors. He also oversees the content delivery of training courses, such as Online and Social Media Analytics and AI for Communicators ---------- Dmytro Plieshakov who is founder and CEO at Osavul. His work focuses on countering disinformation and influence operations using AI technologies. ---------- LINKS: Report: https://www.osavul.cloud/cases/nato-virtual-manipulation-brief https://www.osavul.cloud/ ---------- SUPPORT THE CHANNEL: https://www.buymeacoffee.com/siliconcurtain https://www.patreon.com/siliconcurtain ---------- TRUSTED CHARITIES ON THE GROUND: kharpp - Reconstruction project supporting communities in Kharkiv and Przemysl https://kharpp.com/ Save Ukraine https://www.saveukraineua.org/ Superhumans - Hospital for war traumas https://superhumans.com/en/ UNBROKEN - Treatment. Prosthesis. Rehabilitation for Ukrainians in Ukraine https://unbroken.org.ua/ Come Back Alive https://savelife.in.ua/en/ Chefs For Ukraine - World Central Kitchen https://wck.org/relief/activation-chefs-for-ukraine Ukrainian Freedom News https://www.ukrainianfreedomnews.com/donation/ UNITED24 - An initiative of President Zelenskyy https://u24.gov.ua/ Serhiy Prytula Charity Foundation https://prytulafoundation.org NGO “Herojam Slava” https://heroiamslava.org/ NOR DOG Animal Rescue https://www.nor-dog.org/home/ ---------- PLATFORMS: Twitter: https://twitter.com/CurtainSilicon Instagram: https://www.instagram.com/siliconcurtain/ Podcast: https://open.spotify.com/show/4thRZj6NO7y93zG11JMtqm Linkedin: https://www.linkedin.com/in/finkjonathan/ Patreon: https://www.patreon.com/siliconcurtain ---------- Welcome to the Silicon Curtain podcast. Please like and subscribe if you like the content we produce. It will really help to increase the popularity of our content in YouTube's algorithm. Our material is now being made available on popular podcasting platforms as well, such as Spotify and Apple Podcasts.
(3:40) - The Secret to Super-Fast Robot Vision This episode was brought to you by Murata, the market leader in leading-edge electronics components. Click HERE to learn more about Murata's contributions to the future of autonomous vehicles. Become a founding reader of our newsletter: read.thenextbyte.com As always, you can find these and other interesting & impactful engineering articles on Wevolver.com.
Today's episode features Dr. Tara Akhavan, Global Innovation and Ecosystems Director at Forvia. Dr. Akhavan is an award-winning technology entrepreneur with over 15 years of experience in leadership and product strategy. She is the Founder, President & CEO of Faurecia, a groundbreaking company in the field of Perceptual Display/Image Processing for both consumer and automotive markets. Faurecia Irystec was acquired by Faurecia (Forvia) in 2020, where Tara now serves as the Global Innovation & Ecosystems Director, overseeing startups internally and externally, as well as leading the central Innovation team. She holds a Ph.D. in Image Processing and Computer Vision and actively contributes to industry committees such as CIE and SID. In our conversation, Dr. Akhavan unveils the fascinating journey from her Vienna PhD to pioneering human-centric mobility, teasing exclusive details about Forvia's CES 2024 presence and sharing insights on the impactful intersection of Emotion AI and innovative automotive technologies. Links of interest:[Watch] Tara's TED Talk: Technology Followers or Leaders? | TEDxMontrealWomen - https://www.youtube.com/watch?v=AFnNCuylXR4[News] Consumer Electronics Show : 4 FORVIA technologies win CES 2024 innovation awards: https://www.forvia.com/newsroom/consumer-electronics-show-4-forvia-technologies-win-ces-2024-innovation-awards[News] Smart Eye Announces Innovative Emotion Generative AI Capability and Key Partnerships at CES 2024: https://www.businesswire.com/news/home/20240104295374/en/Smart-Eye-Announces-Innovative-Emotion-Generative-AI-Capability-and-Key-Partnerships-at-CES-2024 [More News] FORVIA and Smart Eye join forces on groundbreaking emotion AI demo at CES 2024: https://www.faurecia.com/en/newsroom/forvia-and-smart-eye-join-forces-groundbreaking-emotion-ai-demo-ces-2024 [Learn More] About Forvia: https://www.forvia.com/[About] Facial Expression Analysis Emotion AI: https://smarteye.se/technology/emotion-ai-facial-expression-analysis/
It's already been one year of JWST operations. In this episode we accumulated all the major science results, all the amazing images and graphs, all the important discoveries and controversies. Enjoy the ULTIMATE GUIDE to the first year of James Webb.
Unlock the power of Data Augmentation in our latest episode of 'The AI Frontier.' We delve into how this innovative technique can creatively increase your dataset size, enhancing your machine learning models' performance. From understanding the concept, exploring various techniques for images, text, and audio, to discussing advanced methods like GANs and autoencoders, this episode is a comprehensive guide to data augmentation. Tune in to discover how to leverage data augmentation in your AI projects and boost your model's efficiency.Support the Show.Keep AI insights flowing – become a supporter of the show!Click the link for details
Segment 1 : who is Hatem? Segment 2 : AI & NLP in Africa are the next disruptors Segment 3 : AI & NLP applications in Africa for equity, inclusion and diversity Bio : Hatem Haddad received a doctorate in Computer Science and Information Systems from University Grenoble Alpes, France. He was a Postdoctoral Fellow at VTT Technical Research Center of Finland and at Norwegian University of Science and Technology. He occupied assistant professor positions at Grenoble Alpes university (France), at UAEU (EAU), at Sousse university (Tunisia), at Mevlana university (Turkey) and at ULB (Belgium). He worked for industrial corporations in R&D at VTT Technical Research Centre of Finland and Institute for Infocomm Research, Image Processing and Applications Lab of Singapore. He was an invited researcher at Leibniz-Fachhochschule School of Business (Germany) and Polytechnic Institute of Coimbra (Portugal). His current research interests include Natural Language Processing, Machine Learning and Deep Learning. He is a Program Chair in various global conferences and serves as a reviewer for relevant journals and conferences in the Artificial Intelligence field. He is a board member of Masakhane and WiNLP (Widening NLP). --- Send in a voice message: https://podcasters.spotify.com/pod/show/mediterranean-sustainable/message
Will Bachman is hosting a conversation with Jamylle Carter, a member of the Harvard and Radcliffe class of 1992. Jamylle graduated from Harvard with a degree in mathematics and went on to complete a PhD in math from UCLA. After UCLA, she also held a postdoctoral fellowship at the Institute of Mathematics and its Applications in the University of Minnesota. Math Think Tanks and Researching Algorithms Will and Jamylle had a conversation about math think tanks, in which Jamylle described her experience. She had spent four years in a math think tank at a university, and then another four years at a Mathematical Sciences Research Institute in Berkeley. Jamylle fell in love with the Bay Area and decided to stay, working as an adjunct professor, running a Math Circle and working at a science museum in San Francisco. She has been a math professor at Diablo Valley College since 2009. Jamylle explained that a math think tank is a place funded by the National Science Foundation and other private funders for mathematicians to leave their university appointments and focus on their research with other people in the field. Jamylle's research was applied math, and it was for image processing. Jamylle was researching a new algorithm to solve a mathematical problem. The problem was related to blurring or noise in an image, and the goal was to approximate the original picture as closely as possible. She was looking at optimization methods, which are mathematical techniques that can find the best answer to a problem. Teaching Math in the Math Circle Program Jamylle then went on to explain a math circle she organized for middle school kids. The math circle was inspired by a program from Eastern Europe that was designed to expose kids to higher level math. The math circle would also give kids a chance to struggle with a problem and get excited about learning. Jamylle learned more about the program while at the Mathematical Sciences Research Institute. In 2007-2008, the MSRI sponsored a Berkeley Math Circle for the children of professors and wealthy families. Seeing the need for a Math Circle for a different demographic, Jamylle proposed an Oakland Math Circle for black middle school students in order to challenge the idea that black people can't like math. With the help of the Exploratorium, the Institute for the Advanced Study of Black Family Life and Culture, and the Museum of African American Technology, she was able to get grants totaling $8,000 to run the Oakland Math Circle. The Math Circle ran hands-on activities to engage students, teaching them that it was okay to be both black and like math. She taught all the modules, did the recruitment and covered topics such as rocket science, probability and music and math. She also mentioned that in minoritized spaces, students tend to work alone, which can be due to racism, fear of being too nerdy, or fear that they won't be seen as capable due to their race. Carter's motivation for starting the Math Circle program was to provide a space where Black kids could come together and enjoy doing math without such pressures. Deborah Hughes Hallett was a Harvard graduate student who eventually became faculty in the Harvard Math Department. She never earned her doctorate, and unfortunately the math department still never treated her like real faculty. She was also at the forefront of the calculus reform movement and wrote textbooks to help teach the subject. Deb was an important figure in the Math Department, providing guidance and support to students in need and advocating for reform in the subject. Designing a New Math Curriculum Jamylle and Will discussed math education and its importance. Will then asked about how to build a high school math curriculum from the ground up. Jamylle believes that everyone should have a solid background in arithmetic, such as fractions, decimals, and percentages. She also suggests introducing courses on probability, statistics, computer science, and using spreadsheets. Additionally, she wants to focus on dimensional analysis and number sense in order to help students become more comfortable with math. Finally, she believes it's equally important to teach math in a way that avoids creating math phobias and traumas, and to make sure teachers are supported and not overworked. Jamylle talks about the courses and professors that have had an impact on her life. Carter starts by talking about her undergraduate advisor Deborah Hughes Hallett, who she credits with helping her to graduate. A class that stood out was music 51, a year long music theory course. Timestamps 08:24 Investigating Mathematical Methods for Image Optimization 11:15 Analysis of Total Variation Method for Image Processing 15:50 Optimization Problems and Finite Time Solutions 18:13 Image Enhancement Technology 20:00 Math Circles and Problem Solving 31:46 Intersecting Racial Identity and Mathematical Affinity 40:13 Comparing Experiences at Harvard and a Historically Black College 48:19 Supportive Mentorship in the Math Department 52:58 Music Theory and Performance 1:00:59 Exploring Equity in Mathematics Education Links: https://mathematicallygiftedandblack.com/honorees/jamylle-carter/ CONTACT INFO: https://twitter.com/CarterJamylle https://www.linkedin.com/in/jamylle-carter-3184259/
I needed to clean up my Lunar Eclipse shot recently, so I gave Topaz Photo AI a try and was highly impressed - DNGs in Capture One Pro also work now! Details on blog: https://mbp.ac/795 Music by Martin Bailey
I needed to clean up my Lunar Eclipse shot recently, so I gave Topaz Photo AI a try and was highly impressed - DNGs in Capture One Pro also work now! Details on blog: https://mbp.ac/795 Music by Martin Bailey
In this episode I have covered top 3 programming languages used for Image Processing Open CV Python Matlab which is better Application in Industry check the episode on various platform https://www.instagram.com/podcasteramit Apple :https://podcasts.apple.com/us/podcast/id1544510362 Huhopper Platform :https://hubhopper.com/podcast/tech-stories/318515 Amazon: https://music.amazon.com/podcasts/2fdb5c45-2016-459e-ba6a-3cbae5a1fa4d Spotify :https://open.spotify.com/show/2GhCrAjQuVMFYBq8GbLbwa
This episode explains you the concept of digital image processing in world simplest and easiest way Analog vs Digital Image vs Digital Image Types of Image Different Image Formats Types of Filters Application of Image Processing in Different Fields check the episode on various platform https://www.instagram.com/podcasteramit Apple :https://podcasts.apple.com/us/podcast/id1544510362 Huhopper Platform :https://hubhopper.com/podcast/tech-stories/318515 Amazon: https://music.amazon.com/podcasts/2fdb5c45-2016-459e-ba6a-3cbae5a1fa4d Spotify :https://open.spotify.com/show/2GhCrAjQuVMFYBq8GbLbwa
An Image Processing approach to identify solar plages observed at 393 37 nm by Kodaikanal Solar Observatory by Sarvesh Gharat et al. on Thursday 22 September Solar Plages are bright chromospheric features observed in Ca II K photographic observations of the sun. These are regions of high magnetic field concentration thus tracer of magnetic activity of the Sun and are one of the most important features to study long-term variability of the Sun as Ca II K spectroheliograms are recorded for more than a century. . However, detection of the plages from century-long databases is a non-trivial task and need significant human resources for doing it manually. Hence, in this study, we propose an image processing algorithm that can identify solar plages from Ca II K photographic observations. The proposed study has been implemented on archival data from Kodaikanal Solar Observatory. To ensure that the algorithm works, irrespective of noise level, brightness, and other image properties, we randomly draw a sample of images from the data archive to test our algorithm. arXiv: http://arxiv.org/abs/http://arxiv.org/abs/2209.10631v1
सुनिए अमित भट्ट के जीवन की प्रेरक कहानी। डेटा साइंटिस्ट से कॉर्पोरेट ट्रेनर बनने तक का ये सफर अमित भट्ट ने बड़ी ही खूबसूरती से तय किया है। ये अबतक कई आईटी एंड सर्विस इंडस्ट्री में काम कर चुके है और आज एक अनुभवी कॉर्पोरेट ट्रेनर है, जो पेशेवरों को उनके व्यावसायिक विचारों का उपयोग करने और उन्हें क्रियान्वित करने में उनकी मदद करते है। इसके अलावा ये एक स्टोरी टेलर एंड पॉडकास्टर भी है। जी हाँ अमित अपनी नौकरी के अलावा टेक्निकल स्टोरीज को शौक के तौर पर पॉडकास्ट करते हैं। ये डिजिटलीकरण पर विशेष ज़ोर देते है और हर किसी को इसे अपनाने के लिए प्रेरित करते है। अपने इतने वर्षों के अनुभव के आधार पर इनका मानना है कि किसी क्षेत्र विशेष में डिग्री हासिल करने से ज़्यादा बेहतर होता है अपनी स्किल्स को बेहतर बनाना। इसी मन्त्र को फॉलो करते हुए इन्होंने आज सफलता का ये मुकाम हासिल किया है। पूरी कहानी पढ़ें: https://stories.workmob.com/amit-bhatt-technology वर्कमोब द्वारा #मेरीकहानी कार्यक्रम के माध्यम से एक नयी पहल शुरू की गयी है जिसके ज़रिये हर कोई छोटे बड़े बिज़नेस ओनर्स अपनी प्रेरक कहानियों को यहाँ सभी के साथ साझा कर सकते है। क्योंकि हर शख्स की कहानी में है वो बात जो जीवन को बदलकर एक नयी दिशा दिखाएगी, और ज़िन्दगी में ले आएगी आशा की एक नयी चमकती किरण। #बनाओअपनीपहचान #प्रेरककहानियाँ #अमितभट्ट #डेटासाइंटिस्ट #कॉर्पोरेटट्रेनर #आईटीएंडसर्विसइंडस्ट्री #स्टोरीटेलर #पॉडकास्टर #टेक्निकलस्टोरीज जानिए वर्कमोब के बारे में: जुड़िये वर्कमोब पर अपनी कहानी साझा करने और प्रेरणादायक कहानियाँ देखने के लिए। ये एक ऐसा मंच है जहां आप पेशेवरों, लघु व्यापारियों, उद्यमियों और सामाजिक कार्यकर्ताओं की वीडियो कहानियां देख सकते हैं और दूसरों को प्रेरित करने के लिए अपनी व्यक्तिगत और व्यावसायिक कहानी सभी के साथ साझा कर सकते हैं। आपकी कहानी में लोगों को आशा देने, प्रेरणा देने और दूसरों का जीवन बदलने में मदद करने की एक अद्भुत क्षमता है। यह 100% मुफ़्त है। इस लिंक पर क्लिक करें और देखें प्रेरक कहानियां https://stories.workmob.com/ हमारे ऐप्प को डाउनलोड करें: Android: https://play.google.com/store/apps/details?id=com.workmob iOS: https://apps.apple.com/in/app/workmob/id901802570
Today's guest is Gil Perry, Co-Founder and CEO of D-ID the technology company behind the viral MyHeritage clips where old photos have been brought to life.A former Commander in the Special Forces unit of the Israeli Defence Forces, Gil and his army friends and co-founders started D-ID to work on protecting images and videos from facial recognition software but not long after and when Covid came, they pivoted their startup to focus on creating a suite of what they call Creative Reality tools that use their deep learning algorithms, image processing and neural networks to create high quality videos from still images. If you haven't tried it yet, head to the MyHeritage site and give it a go - I can only describe the experience as magical.I have long-been interested in how we can preserve the memory of our loved ones so Gil and I talk about this, as well as holograms, deepfakes, how and why the company pivoted, what lesson Gil took from the special forces to entrepreneurship and, of course, D-ID's incredible work – from bringing history to life, right through to the metaverse.Enjoy!Gil on LinkedIn / D-ID website / Twitter Danielle Twitter / Instagram / Newsletter-----Mentioned in the episode:MyHeritage Deep NostalgiaHologram of Robert Kardashian (gift from Kanye West to Kim Kardashian)Glimpse Group
Do you know the difference between creating a class instance and initializing it? Would you like an interactive tour of the Python Pillow library? This week on the show, Christopher Trudeau is here, and he's brought another batch of PyCoder's Weekly articles and projects.
For content creators, one of the biggest issues can be ensuring that what the end consumer hears and sees is exactly what you intended. In theaters, this problem has been mostly remedied by creating industry standards that apply across theater chains and projector manufacturers. However, as more media has moved to home viewing, the experience … Continue reading Pixelworks: 20-year leader in image processing innovation @ CES 2022 → The post Pixelworks: 20-year leader in image processing innovation @ CES 2022 appeared first on Tech Podcast Network.
For content creators, one of the biggest issues can be ensuring that what the end consumer hears and sees is exactly what you intended. In theaters, this problem has been mostly remedied by creating industry standards that apply across theater chains and projector manufacturers. However, as more media has moved to home viewing, the experience changes significantly, depending on the hardware being used - particularly the television. This is exactly the problem that Pixelworks is looking to solve with TrueCut Motion.Who is Pixelworks?Pixelworks is a 20-year leader in image processing innovation, providing solutions and technology that enable highly authentic viewing experiences with superior visual quality. Their products are used by some of the biggest names in the consumer electronics, professional displays, and video streaming industries. The company is committed to delivering even more impressive image processing technologies in the years to come!What is TrueCut Motion?TrueCut Motion is Pixelworks' new technology designed with the needs of content creators in mind. The goal is to ensure that motion picture quality and sound are delivered exactly as they were intended, regardless of which display or projector a viewer chooses to use. The company has created multiple levels of TrueCut Motion technology, allowing them to provide solutions for any budget or need.As we move further into the era of streaming media, it's more important than ever to have a technology provider that understands the challenges and intricacies of delivering high-quality content. Pixelworks is that company, and TrueCut Motion is their solution for ensuring an optimal viewing experience for everyone, regardless of their setup.How does TrueCut Motion work?TrueCut Motion technology has been developed over Pixelworks' 20-year history and is an important part of their continued innovation. It works by manipulating pixels to adjust for perceived judder, which can make content feel choppy or unnatural when viewed on a television. The team has worked diligently to create algorithms that fully compensate for this, delivering a smooth and seamless picture.The Pixelworks team has also worked on noise reduction, eliminating unwanted visual artifacts that could otherwise detract from the viewing experience. The company incorporates three different levels of noise reduction into TrueCut Motion technology - Ultra Precision, Precision, and High Efficiency - all depending on the budget or needs of their customer. They are the only company to offer this level of customization for noise reduction.Pixelworks has big plans for TrueCut Motion and image processing in general. They want to continue developing technologies that will make it easier than ever for content creators to deliver an optimal viewing experience across all screens - from smartphones and tablets to large home theater displays. Pixelworks is also working on ensuring that their technology works with newer display formats like HDR, making it even more impressive and useful for content creators across all media!TrueCut Motion and TCLAt CES this year, Pixelworks announced a new partnership with TCL - one of the world's largest television manufacturers. This means that the TrueCut Motion technology will be integrated into select TCL televisions, delivering an even better viewing experience for consumers. With Pixelworks and TCL working together, we're sure to see some impressive advancements in the near future!SummaryIf you want more information about Pixelworks, TrueCut Motion, or their partnership with TCL check out the Pixelworks website.Interview by Todd Cochrane of Geek News Central.Sponsored by: Get $5 to protect your credit card information online with Privacy. Amazon Prime gives you more than just free shipping. Get free music, TV shows, movies, videogames and more. The most flexible tools for podcasting. Get a 30 day free trial of storage and statistics.
For content creators, one of the biggest issues can be ensuring that what the end consumer hears and sees is exactly what you intended. In theaters, this problem has been mostly remedied by creating industry standards that apply across theater chains and projector manufacturers. However, as more media has moved to home viewing, the experience changes significantly, depending on the hardware being used - particularly the television. This is exactly the problem that Pixelworks is looking to solve with TrueCut Motion.Who is Pixelworks?Pixelworks is a 20-year leader in image processing innovation, providing solutions and technology that enable highly authentic viewing experiences with superior visual quality. Their products are used by some of the biggest names in the consumer electronics, professional displays, and video streaming industries. The company is committed to delivering even more impressive image processing technologies in the years to come!What is TrueCut Motion?TrueCut Motion is Pixelworks' new technology designed with the needs of content creators in mind. The goal is to ensure that motion picture quality and sound are delivered exactly as they were intended, regardless of which display or projector a viewer chooses to use. The company has created multiple levels of TrueCut Motion technology, allowing them to provide solutions for any budget or need.As we move further into the era of streaming media, it's more important than ever to have a technology provider that understands the challenges and intricacies of delivering high-quality content. Pixelworks is that company, and TrueCut Motion is their solution for ensuring an optimal viewing experience for everyone, regardless of their setup.How does TrueCut Motion work?TrueCut Motion technology has been developed over Pixelworks' 20-year history and is an important part of their continued innovation. It works by manipulating pixels to adjust for perceived judder, which can make content feel choppy or unnatural when viewed on a television. The team has worked diligently to create algorithms that fully compensate for this, delivering a smooth and seamless picture.The Pixelworks team has also worked on noise reduction, eliminating unwanted visual artifacts that could otherwise detract from the viewing experience. The company incorporates three different levels of noise reduction into TrueCut Motion technology - Ultra Precision, Precision, and High Efficiency - all depending on the budget or needs of their customer. They are the only company to offer this level of customization for noise reduction.Pixelworks has big plans for TrueCut Motion and image processing in general. They want to continue developing technologies that will make it easier than ever for content creators to deliver an optimal viewing experience across all screens - from smartphones and tablets to large home theater displays. Pixelworks is also working on ensuring that their technology works with newer display formats like HDR, making it even more impressive and useful for content creators across all media!TrueCut Motion and TCLAt CES this year, Pixelworks announced a new partnership with TCL - one of the world's largest television manufacturers. This means that the TrueCut Motion technology will be integrated into select TCL televisions, delivering an even better viewing experience for consumers. With Pixelworks and TCL working together, we're sure to see some impressive advancements in the near future!SummaryIf you want more information about Pixelworks, TrueCut Motion, or their partnership with TCL check out the Pixelworks website.Interview by Todd Cochrane of Geek News Central.Sponsored by: Get $5 to protect your credit card information online with Privacy. Amazon Prime gives you more than just free shipping. Get free music, TV shows, movies, videogames and more. The most flexible tools for podcasting. Get a 30 day free trial of storage and statistics.
Wepa! I'm Marina. I am a technologist, mom, podcast host, leadership coach, cruciverbalist and aquarian. ;) UNBOSSED is “Stories of Amazing Women in Chicago”. If you are a new listener to UNBOSSED, we would love to hear from you. Please visit our Contact Page and let us know how we can help you today! Support this podcast: https://anchor.fm/marina-malaguti In this episode: Interview with Geeta Pyne, Chief Enterprise Architect - Platform and Architecture at Intuit. Geeta Pyne has extensive experience in Technology strategy,architecture and digital solution delivery. As lead architect she has led global companies through digital transformations, cloud migrations and enabled new business models. She has worked in both Products development and Business Technology solution delivery. She brings enterprise mindset while working in a startup, agile fashion. Recently as the Vice President of Data at Zuora she drove data strategy and execution to help the company scale to $1 Billion. Prior to Zuora, she worked at multiple high-tech companies from healthcare to supply chain to cloud and data management. She started her career as a research scientist developing algorithms for Satellite Image Processing and takes pride in holding IP in India's first parallel computer PARAM and Image Processing system ISROVISION. In the recent past she has laser focused on data, machine learning, AI and using data as a competitive differentiator. Key High|lights: Woman in tech growing up in India Immigrant data leader India Space Organization Getting into the room Horizontal Career Growth and Vertical Career Growth Memorable Quotes: “If you are good at something, dig deep into your craft”- Geeta Pyne, Chief Enterprise Architect at Intuit “Some people go for opportunities, others go for possibilities. Geeta goes for possibilities.” - One of Geeta's managers “Don't only think about who you invite [into a meeting], also think about who you are excluding.” - Geeta Pyne, Chief Enterprise Architect at Intuit “Nobody messes with Geeta Pyne” - Geeta Pyne, Chief Enterprise Architect at Intuit Useful Links and Resources: https://www.linkedin.com/in/geetapyne/ https://tipco.medium.com/lessons-from-an-immigrant-data-leader-geeta-pyne-chief-enterprise-architect-intuit-544f039d09c6 Join the Conversation Our favorite part of recording a live podcast each week is participating in the great conversations that happen on our live chat, on social media, and in our comments section. Follow our Podcast Spotify: https://lnkd.in/eUhfH8E Apple Podcasts: https://lnkd.in/e7cWtBv Google Podcasts: https://lnkd.in/enjChPt Youtube: https://www.youtube.com/channel/UCDTz6_FepG04QTs1BjFLBjw/ And all others... --- Support this podcast: https://anchor.fm/marina-malaguti/support
This is the hundred and fifty-second episode of the GameDev.tv Community Podcast. Nathaniel is a Image Processing engineer, Georgia Tech PHD graduate, machine vision expert, and Director of Software at EZ Automation Systems. Hear his advice for young programmers, stories of his childhood and how he came to be, and of course how he used the frequency of sound waves to estimate capacitance of florescent wire.https://www.gamedev.tv/p/low-poly-characters/?product_id=3486107&coupon_code=K-B&affcode=45216_dezckag6Landscapes:https://www.gamedev.tv/p/low-poly-landscapes/?product_id=1319848&coupon_code=K-B&affcode=45216_dezckag6C++ Fundamental:https://www.gamedev.tv/p/cpp-fundamentals/?product_id=3266750&coupon_code=K-B&affcode=45216_dezckag6Unity Mobile:https://www.gamedev.tv/p/unity-mobile/?product_id=3190328 &coupon_code=K-B&affcode=45216_dezckag6Unreal C++:https://www.gamedev.tv/p/unreal-engine-c-developer-4-22-learn-c-and-make-video-games/?product_id=1319848&coupon_code=K-B&affcode=45216_dezckag6Unity 3D:https://courses.gamedev.tv/p/complete-unity-developer-3d/?product_id=1319848&coupon_code=K-B&affcode=45216_dezckag6Unity 2D:https://courses.gamedev.tv/p/complete-unity-developer-2d/?product_id=1319848&coupon_code=K-B&affcode=45216_dezckag6Unity RPG Core Combat:https://courses.gamedev.tv/p/unity-rpg/?product_id=1319848&coupon_code=K-B&affcode=45216_dezckag6Math For Games:https://courses.gamedev.tv/p/math-for-games/?product_id=1319848&coupon_code=K-B&affcode=45216_dezckag6Enjoy the Podcast!Support the show (https://www.gamedev.tv/p/complete-unity-developer-3d/?product_id=1319848&coupon_code=The_K_B&affcode=45216_dezckag6)
Dr. Paul Yi interviews Bhavik Patel Senior Associate Consultant Associate Professor Department of Radiology Director of AI on recent article Impact of Upstream Medical Image Processing on Downstream Performance of a Head CT Triage Neural Network. Hooper et al. Radiology: Artificial Intelligence 2021; 3(4):e200229 Introduction to the podcast by Dania Daye, MD, PhD
Geeta Pyne is the Chief Enterprise Architect - Platform and Architecture at Intuit (NASDAQ:INTU). Intuit is a finance and accounting software development company serving millions of customers worldwide with popular financial products like TurboTax, QuickBooks, Credit Karma, and Mint. The company has 10,000 employees worldwide with $9.6billion in revenue in 2021 Geeta is currently leading Capability driven Architecture (aka City Map) and Design to drive Intuit's Platform Velocity by delivering design patterns, composable services, and reusable solutions. Geeta is also a Board Member of SIM San Francisco Bay Area. The Society for Information Management (SIM) is the largest National technology leadership organization in the US providing Programs, Events, Employee Development, and Networking for members. Before Intuit, Geeta was the Vice President of Data at Zuora, where she drove data strategy, and architecture to build the foundation of data-driven decision making and execution to help the company scale to $1 Billion. With specialization in SaaS, data, analytics, hybrid cloud, digital transformation, cloud migrations, and enabling new business models, Geeta has been working in the world of data, technology, and engineering for almost two decades in verticals ranging from healthcare to supply chain to cloud and data management. Prior to Zuora, she worked in multiple senior leadership roles at high-tech companies, including Veritas Technologies, VMware, BMC Software, and Mount Sinai Health System Geeta also holds an IP in India's first parallel computer PARAM and Image Processing system ISROVISION. This exclusive episode starts with Geeta sharing about the early days of her career at the Indian Space Research Organisation, and how she came to patent India's first parallel computer and image processing system. How the lack of modern cloud computing meant it was essential for us to fully utilise what we had in the old days. The role of enterprise architecture, especially in the context of data science. The importance of thinking about the big picture and building reusable blocks so that they can be used across the company and multiple products. The role her team plays and how they create the enterprise architecture culture with a focus on reusable blocks that allows Intuit to continue building itself as a platform company. Geeta also shares many examples of AI, ML models and emerging technologies being used at Intuit to solve the problems for their customers. Whether you're a technology or a non-technology company, as Geeta puts it, necessity is the mother of invention and all of us are going to need software to empower what we do in the business. If you're a data scientist or business executive, tune in to listen to Geeta Pyne on this episode to understand how enterprise architecture and reusable blocks are equally essential in data science projects. If you have any questions for Geeta or myself, make sure you send us an email or a message on LinkedIn. This episode is sponsored by the new program at DDA. It's an analytics leader mentorship program for senior managers and executives in the business team who want to develop a data-driven business to drive customer experience excellence. For a small one-off annual fee, you get to book Unlimited Strategy Sessions for a Full Year. For more information about this program, please reach out to DDA! #BusinessAnalytics #MachineLearning #NaturalLanguageProcessing #Agile #CustomerExperience #DataScience --- Send in a voice message: https://anchor.fm/analyticsshow/message
The following is my conversation with Abhi Aiyer and Ward Peeters, two lead engineers behind Gatsby Cloud and the recently announced Gatsby v4, which is at the forefront of what I think is the most significant change in the Jamstack landscape in the past 2 years.Watch the video version here. Links: Gatsby 4 Netlify DPR My blogpost on Smart Clients vs Smart Servers Timestamps: [00:00:00] Cold Open [00:00:28] Swyx Intro [00:01:59] Call Start [00:03:07] Gatsby v4 [00:06:23] Incremental Builds [00:07:16] Cache Invalidation [00:09:03] Gatsby DSG vs Netlify DPR [00:09:35] Abandoning Redux for LMDB [00:11:50] Parallel Queries (PQR) [00:13:32] Gatsby DSG [00:15:24] Netlify DPR vs Gatsby DSG [00:19:19] The End of Jamstack [00:22:12] Tradeoffs and Performance [00:24:34] Image Processing [00:27:25] Automatic DSG [00:29:33] Gatsby Cloud vs Netlify [00:33:34] Gatsby vs Next.js [00:35:41] Gatsby and the Content Mesh [00:37:19] React 18 and Gatsby [00:39:45] Custom rendering page fragments with React 18 [00:42:10] Server Components in Limbo [00:43:33] Smart Servers vs Smart Clients [00:45:21] Apollo and Open Source Startup Strategy [00:47:06] TMA: Too Many Acronyms [00:49:16] Gatsby for Docs Transcript [00:00:00] Cold Open [00:00:00] Abhi Aiyer: And so with LMDB in place, right? We have workers that can read and write to LMDB, which allows us to run parallel queries. So PQR was a huge advancement for us. I think we saw up to like 40% reduction in query running time. And build times went down. We had a goal, I think it was like, we'd try to look for at least 20% reduction in build times and I think we hit 26%, so all cool wins, you know? [00:00:28] Swyx Intro [00:00:28] swyx: The following is my conversation with Abhi Aiyer, and Ward Peeters, two lead engineers behind Gatsby Cloud, and the recently announced Gatsby V4, which is at the forefront of what I think is the most significant change in the JAMstack landscape in the past two years. We discussed how parallel query writing PQR and deferred static generation DSG are achieving 40% faster queries and 300% faster overall builds. [00:00:53] And they did a wonderful job handling the most impolite questions I could think of, including whether it Gatsby Cloud is a Netlify clone or the Gatsby should just be a data layer on top of Next.js and how they're dealing with TMA too many acronyms in web development. This conversation should be viewed together with my past discussions, with Sunil Pai and Misko Hevery in considering the cutting-edge of web development today. Online discussions often present a binary split in that your technical choices either have to optimize for developer experience or user experience. [00:01:25] But I find that it is builders like Abhi and Ward and Misko and Sunil who are constantly trying to improve the experience of developers in building great user experiences by default. I hope you enjoy these long form conversations I'm trying to produce with amazing developers. I still don't have a name for it. [00:01:41] And I still don't know what the plan is. I just know that I really enjoy it. And the feedback from you guys have been really great. So if you like this, share with a friend, if you have other requests for guests, tag them on social media, I basically like to make this a space where passionate builders and doers can talk about their craft and where things are going. [00:01:58] So here's the interview. [00:01:59] Call Start [00:01:59] Abhi Aiyer: I'm Abhi Aiyer. I'm a principal engineer at Gatsby. Thanks for having us. [00:02:05] Ward Peeters: My name is Ward Peeters. I'm a staff software engineer at Gatsby and I'm from Belgium. And I've been working mostly on the open source side. [00:02:15] Abhi Aiyer: I forgot to say where I'm from. I'm from Los Angeles, you know, Hollywood, [00:02:21] swyx: I'm actually heading down to LA, [00:02:22] Abhi Aiyer: in a couple of weeks, there's, [00:02:24] swyx: I'm going to Kubecon, which is like a very interesting thing for a front end engineer to end up at. But that's where my career has taken me. [00:02:34] So this conversation started because I had a chat with Sunil, on this podcast that I accidentally launched. I don't think we did Gatsby much, a good favor. [00:02:45] Like we both saw the new updates and I didn't get to say the nice things that I thought about Gatsby. I should also say that I used to have my blog on Gatsby and I no longer do. I used to work at Netlify and I no longer do. There's a lot of history here for me with Gatsby. It's been a while since I caught up, and I'm curious to see or get the latest. [00:03:07] Gatsby v4 [00:03:07] swyx: Maybe we should start off with like a quick summary of what's new with Gatsby with Gatsby V4, right? [00:03:13] Abhi Aiyer: Is that a good place to start? Yeah, I think so. [00:03:17] swyx: So first of all, I think the marketing was really nice. Gatsby camp, it seems like a really big push and qualitatively very different from Gatsby 3. Tell me about what the behind the scenes was like. [00:03:30] Abhi Aiyer: Yeah, it was, we're getting better at the marketing side of what we're doing these days and Gatsby 4 was a big push. It really changed how we approach the framework as a whole. [00:03:43] For those who don't know, traditionally Gatsby was a static site generator, purely static. We hold ourselves high on our connections to a content management system. [00:03:55] And we provide a really good data layer there, that takes all those requests that you would normally make to a content manager system, turns them into a, like a store of data that you can then use and query from graph QL. And the big thing that we were hitting before gas before was. Company was growing. [00:04:17] And as more customers were using Gatsby cloud, we started realizing that we couldn't scale to really large sites and large sites is like a misnomer. Like you could be, you could be a 50,000 page site and be considered large given the data that you may have. But we're talking like hundreds of thousands of pages. [00:04:38] And the thing that we kind of realized is not all pages are created equal on your site. Especially the ones from like 20, 15, 20 14, where, you know, no one's looking at that people, those pieces of content, if you're a site with a huge archive of content yeah, you should probably go check Google analytics to see how, you know, how, how many people are actually viewing your pages. [00:05:02] And the way gets me. And we'll get into this a little bit later, but today Gatsby isn't as smart as it should be in what pages should be rebuilt. For example, we're looking at the gatsbyjs.com page right here, but there are other data dependencies. This pure content. Like if you look at the nav bar, there's a whole bunch of items there. [00:05:22] And we have this navbar bar on all of our pages, cause that's what a website has, but the problem with Gatsby today and will be changed in the future is. If I change anything about this nav bar, any page, that depends on the nav bar now has a dependency that needs to be invalidated. [00:05:39] And so now I have a hundred thousand pages with this navbar I have 200,000 pages of this nav bar and I spelled Gastby instead of Gatsby or something, the navbar I made a typo and now I'm paying for. A hundred thousand pages of reload to rebuild. And so we just saw that and that this happens a lot, right? [00:05:57] This is a very small example, but this happens a lot to our customers and building a hundred thousand pages is not necessarily easy thing to do. There's memory requirements that come in. There is, what kind of machine are you doing this thing on? And so we had to figure out a way to scale Gatsby and do things differently. [00:06:15] We were traditionally static and now we're trying to be somewhere in between, you can go dynamic or you could go static and it's up to you. [00:06:23] Incremental Builds [00:06:23] swyx: So the new rendering options are SSG, DSG and SSR. Is ISR gone? [00:06:32] Ward Peeters: Well, that's what Next.js has is doing. And I'm like, wait, [00:06:36] swyx: we never have guessed. [00:06:38] We had an incremental mode. [00:06:41] Abhi Aiyer: What do you [00:06:41] Ward Peeters: call it? Yes. And that's still all statically. So when we do it, we have it in open source where we in V3 we enabled it where we only build HTML what's necessary on cloud. We have a more pumped up version of it where When you get the data change, we only update that page more rapidly than in open source, but still when you change your navbar with what Abhi said, you still have to rebuild everything because all the pages get invalidated. [00:07:09] So incremental builds works for data updates, but not so much for code changes. [00:07:16] Cache Invalidation [00:07:16] swyx: Right. Okay. Well, I guess, how do you solve cache invalidation? [00:07:26] Ward Peeters: Well, the thing is that because Gatsby owns the data, like the heads of data layer and a source plugins like WordPress, when we source data and to give us a webhook or, we just go to Wordpress and say like, Hey, what has changed? [00:07:40] Data. I was like, okay, these nodes have changed. Or these pieces, like a poster page has been changed. It gets me knows which node is used where, like, for example, this post is used on this section page. It's used in this article and that's all happening already behind the scenes because graph QL shenanigans. [00:07:59] And that's how we can build incremental builds. So we know, okay. Only these spaces need to be built. And that's also where DSG comes in because as a user, you don't have to care about cache invalidation anymore. Cause it's using the same engine as you were used to with like incremental builds. [00:08:15] When you use SSG and I think that's a major benefit of it, where you as a user, don't really have to care about cache control, because it is difficult to manage on a large scale. Like a lot of corporations just say like every 10 minutes we'll remove the cache because it is difficult to get through when change. [00:08:37] Yeah. [00:08:39] swyx: That's pretty funny. At Netlify, one of the reasons that we constantly talk about for CDN level caching, like people say like, you know, why don't you just enable CDN level caching and then just have a regular server render. One of Matt Billman points that he always makes is that people always turn it off the moment there's a bug, it's like, oh, schedule, call, and turn it off. [00:09:02] And then don't turn it back on again. [00:09:03] Gatsby DSG vs Netlify DPR [00:09:03] swyx: Okay. So let's, let's talk about like, DSG. That's the fancy new one and DPR, right? So maybe we should, is there. Yeah, there's a graphic. Okay. All right. This is new. So first of all, like what was the inspiration? What's the backstory I'm always interested in how these things come about. [00:09:21] Abhi Aiyer: I think we were inspired by DPR a lot, you know? But where we saw the benefit of our approach is our data layer, you know, and it took those many steps even before getting to DSG. [00:09:35] Abandoning Redux for LMDB [00:09:35] Abhi Aiyer: So it started actually in like Gatsby 3.10. We had to redo Gatsby's node store. [00:09:42] So traditionally we were using Redux to persist all these the data that we get from content management systems. And we had a particular customer who could not even persist the cache, like a Gatsby cache between builds, because they had so much data that it would OOM when they try to persist the cache. Right. [00:10:03] So for them, they were running cold builds for every build. Even if you had like a warm cache or you had your pods, you know, we use Kubernetes. So like, if you have your pods up, you're doing like an hour and a half cold build for everything. You could like change the typo and it'd be an hour and a half. [00:10:19] And so from there we were like, We need to reduce peak memory utilization and Redux is not going to help us there. And so we started looking into alternatives. We looked at SQL Lite, we looked at Reddis and we landed on LMDB, which is, Lightning memory, mapped database manager. [00:10:39] It's like a file system DB, which is really cool for us because one, it's pretty fast. It allows you to, to have like a query interface, which is good. You can store more data than available RAM. So for a site like this customer who pretty much is blowing up this pod on every warm build. To try to even have a warm build, we could actually store their data now, which then unlocked warm builds for them. [00:11:05] So an hour and a half, that went to 25 minutes, which is pretty good for them. now we have this thing, now we call it Gatsby DB internally. And so now Gatsby is node store is in LMDB. And the cool thing about LMDB is it's just comprised of a bunch of files. You have a lock file and database files. [00:11:26] And if you have files, that means you can move files around. They don't have to be in one place, right. They could be in storage, they can be in a serverless function. They could be anywhere you, you really want. Right. And so that was step one was we needed to redo the node store. And we did that and memory utilization from a lot of customers went down. Enough to unlock a simple thing as a warm build. [00:11:50] Parallel Queries (PQR) [00:11:50] Abhi Aiyer: So then the second thing that this, these other customers were having was like, wow, it takes so long to query, to run queries. Right. And people have like 25,000, 50,000 queries. And I don't know if they should have those that much, but they do. [00:12:05] Right. They do have that much. And it's a huge part of the build time. Right. A lot of people complained that. You know, Gatsby builds are sometimes slow for large sites and we agree. That's totally true. And so our next foray into like improvement was this thing called parallel queries. Which would allow Gatsby to run chunks of queries at a given time and what PQR in his pool, a diagram of it, you know, query running does take a huge percentage of your builds. [00:12:39] But now we can parallelize that across workers in the Gatsby process. But if you were to do this naively with Redux, like a child process can't write to a JavaScript object in the main process. Right. It's hard to maintain state. There's no easy way to communicate between workers to write state and read it. [00:12:59] And so with LMDB in place, we have workers that can read and write to LMDB, which allows us to run parallel queries. Right. We don't need to do things serially, anymore. So PQR was a huge advancement for us. I think we saw up to like 40% reduction in query running time. And build times went down or we had like a goal, like I think it was like, we'd try to look for at least 20% reduction in build times. [00:13:27] And I think we hit 26%, so all cool wins, you know? [00:13:32] Gatsby DSG [00:13:32] Abhi Aiyer: And so then Ward and I, and the team were all just like thinking like, okay, we have LMDB. We've got PQR. Alright, well really we have a Gatsby data layer that can be accessed from anywhere, right? Cause if you can access it from a worker, you can access it in a serverless function cloud run, you know, on your somewhere, anywhere you spin up your own machine and your own office, if you want it well [00:13:56] swyx: steady coast. [00:13:57] How about that? Like an S3 [00:14:00] Abhi Aiyer: bucket, you put it in an S3 bucket, for sure. You know, like you could put those files there and then retrieve them from wherever you want. And so that's when we started thinking like, okay, we have this information now, what can we do to improve the, the life of our users even more. [00:14:19] And then we started thinking about DPR and like, we saw the approach and we were like, wow, this is exactly what we need, but we have Gatsby's data layer that kind of complicates things, but it's not complicated anymore because we can essentially use the data layer wherever we wants. So I'll let ward kind of go from there on like how DSG came about after these like fundamental pieces. [00:14:42] Ward Peeters: Yeah. So then we looked at like ISR DPR and like what's the difference in both of them. And then we saw like ISR that's where you have a refresh timeout and an hour with, in the latest next, you can also being an endpoint to they're getting validated cache, but it's all manual work. And there were many complaints about it's an index. [00:15:02] It's nothing in Gatsby and they complained about stale data, because what Next.js does is you visit the page and then the next time it will update. So I think it's a refresh or something. Yeah. [00:15:15] swyx: Alright. Alright. We don't have to dig through issues on, on the, on our call, but I just wanted to illustrate the problem. [00:15:24] Ward Peeters: Yeah. [00:15:24] Netlify DPR vs Gatsby DSG [00:15:24] Ward Peeters: And then that's where we took it away and then say, okay, DPR. And then I looked at the spec of DPR, like, okay. Can we use the same name or not? And the problem with DPR was they had Atomic deploys. So every change means blow the whole cache away and do everything new and we were like, what do we have incremental builds from there? We don't want to like invalidate the whole cache. We just want to invalidate the pages that got removed. And there's like a GitHub discussion about it, where I commented as well. [00:15:55] And it felt like they didn't want to change the name. Yep. There you go. [00:16:04] swyx: So you said to me, DPR, doesn't need to be opinionated about if the file is part of the atomic deploy. Can you reiterate why? [00:16:13] Ward Peeters: Yeah, the thing is basically because they mentioned like everyday glory needs to blow the cache away and needs to be fresh. [00:16:20] And for me, like it shouldn't be in a spec like DPR should just say you built pieces at build-time and you build those pieces at runtime. That's basically what I was trying to say. And then because we have incremental builds, we only want to invalidate like five pages, even if you use SSG or DSG, we still want to say if you only changed five pages for evil dates to cache for five pages, I couldn't get that from the spec. [00:16:46] I think that's also because Netlify does it their way, which is totally fine, but then that's why we created a new acronym called DPR. And I think it's also probably explains. What we offer as well, a little bit better too, because it's Deferred Static Generation. It's like lazy SSG, something like that, because that's what we do. [00:17:08] Like you can mark a page as defer and that just means we don't do it at build time, but the first time you hit a request. We rebuild it in like a Lambda, I could use Cloud Run, we build it and then we give the response to a user and then also we save it to disk. So from there on, the second request, it's technically an SSG page. [00:17:29] We store it like you have the CDN cache, but we also have it inside our bucket. Like, your S3 buckets or whatever you want to call it. [00:17:37] Abhi Aiyer: Yeah. We're caching responses, but we're also making that file exist as if it existed at build time. And that's a big distinction for us because what that allows us to do in the future would be like, if nothing changed about the data for the given page, then you don't need to DSG again. [00:17:56] Right. Like if nothing changes for, let's say like there's five builds and build a. Something changed in your data dependencies. So now you have a DSG page and then nothing changed for the next five builds, but a user comes and actually visits that page generates the files. It gets cacheed in our data layer or our files storage layer and on build five because nothing changed. [00:18:24] You're not DSGing. Right. You're not going to go through this process again. And so that's we think is the big thing about DSG. [00:18:31] Yeah. And then I think the extra piece of it is because the date, like you can say it it's a benefit or or a negative point of Gatsby, like we source all the data at the build time. [00:18:41] So even if your APIs go down, even with DSG, you still go to our local database. So debts will never go down. Cause if like your site is down, your database will be down as well, but you, you're not dependent of other API. So let's say GitHub goes down or X go down and you need to get that data. We have it locally in our database, so you're still good to go through, still keep that resilience. [00:19:06] And the security even that you, you used to have with Gatsby, and I think that's a main benefit of the whole datalayer piece of Gatsby and DSG. [00:19:17] Yeah. [00:19:18] swyx: Yeah. Perfect. [00:19:19] The End of Jamstack [00:19:19] swyx: So something I always wonder about like, is this basically the last stage of JAMstack like, I feel like we have explored all possible varieties of rendering. [00:19:30] And this is like the end. This is like, this is it right? Like we have all the options. [00:19:34] Ward Peeters: And now it's mixing them together. It's the next step having been static and on bits of your thesis, SSR. Uh, [00:19:43] swyx: okay. I'll put it this way. Do you think that JAMstack at the end of the day after this, maybe like five-year journey of like, Hey, like a WordPress sucks. [00:19:53] That's everyone moves to static. Right. And then, and then we found like, oh yeah, static. Doesn't scale, big surprise. We were telling you that from the beginning. And now okay. Right. Hybrid. Is that it, like, it was that the Jamstack movement in like a five year period? [00:20:10] Abhi Aiyer: I think it's a yes or no. Like evolution is like, I think we're, you know, we're all coming full circle and I think in engineering, particularly we do the same thing all the time, every 10 years or something. Right. But where DSG came into play is for use cases that we saw, you know, and our customers still prefer static. [00:20:31] So I know we're talking about DSG. Like it's like a great thing and it is, but a lot of our customers prefer static and it's really up to their use case. If you're a small site out of a bunch of top of funnel page, any lag in anything, right? Cause DST is not like instant, right? Like you're doing a runtime build essentially. [00:20:51] Right? So in some cases it could be, you know, it could, it could be a longer response time than what the standards should be. And we have customers that won't DSG anything because they have essentially, most pages are top of funnel or high traffic that they would rather just wait.They don't mind waiting for the performance that they would want. [00:21:11] But we also have customers that have hundreds of thousands of pages, like there's one customer that has like a company handbook or something where like, you can see every employee. And like, if they like dogs and like, you know what I'm saying? Like, Bio's and stuff. And they have a lot of employees worldwide, and there, they can only like before DSG, they can only build their site once a week. [00:21:33] Cause it takes like 24 hours to build. What, and now with DSG, they don't really care about someone who no, one's going to view their profile. No offense to that person, but no one's viewing the non CEO's profile. So then how they can, like, you know, and there are other people that are important too. I'm sure, but like now they can actually, you know, make changes to their site. [00:21:55] You know, we actually had to work with them to make sure that, you know, they can build. I mean, previous to DSG, they can build like, at some cadence that we don't necessarily support, but we help support that. So, so just looking static is still king when it makes sense. For sure. [00:22:12] Tradeoffs and Performance [00:22:12] swyx: I feel like it's a bit scary when you tell people like, okay, you're deferring the build. [00:22:16] And then on the first request, someone's going to build that. It's not going to take that long. Yeah. Right. It's not like it's that bad. I think bottom line is, I think people are very scared whenever you say, like, okay, there's a trade off, but you don't quantify the trade-offs. And then they're like, oh, it's bigger in their mind than it really is. [00:22:37] Ward Peeters: Yeah, I think a big problem with the plugin ecosystem is that it's difficult to, to quantify like what's slow and what's not slow. For example, generating an MDX page is more time-consuming because it has to like get some dependencies make sure that they have bundled together, then use react to render and then render again because it's how the Gatsby plugin, is built right now that takes more time than a simple React renderToString with something. [00:23:07] And I think that's the difficult thing to say like, okay, it's some pages will be instant. Some pages might take a second to build or we'll half a second. [00:23:18] swyx: Yeah. The important thing is that there are not data dependencies that you're waiting on. Right. That's usually the slowest part fetch all the data upfront and then you store it in a LMDB cache. [00:23:28] And that's written to a serverless function or written to I guess your build process or whatever. And then people can render that whenever which I think is great. Like, it should be fairly fast, like we're talking tens of milliseconds difference between like for first render, right? [00:23:44] Like something like that. Like I think, I think when you quantify, like, okay, we're talking tens of milliseconds, not hundreds of milliseconds and not thousands of seconds that really helps me with. Put these things in perspective. [00:23:56] Abhi Aiyer: Yeah. But then, you know, people always find a way to screw it up. So say that like, of [00:24:01] swyx: course. [00:24:01] Yeah. But, but you give a realistic benchmark and then you go like, yeah, for these benchmarks, we tested it like a hundred times or something. The median was this, the P 95 was that. That's it like, I mean, people can't really fault you for not accounting for every use case because no one, no one can, but at least you can give a reasonable basis and say like, [00:24:22] Abhi Aiyer: there's, [00:24:23] swyx: there's an up, there's an upper bound to you know, how bad, how the, the, the trade-off like, you know, when, whenever you miss channels, I like to quantify it basically. [00:24:32] Um, that's a good, that's a good idea. [00:24:34] Image Processing [00:24:34] Abhi Aiyer: And like, one thing to know for DSG is like, your data may be like available and that's cool that that may not be the long pole, but let's say you have a portfolio site that generates 20 different types of images for every image. Now you're getting into image processing at runtime, you know? [00:24:54] And so there, there are ways to kind of not do this properly. Right. And or like, for example, let's say your homepage, I love this example. Your homepage has links like to every other page on your site,and it's all DSG, right? So you load the homepage and because Gatsby does prefetch for link tags are doing Gatsby link to other pages. [00:25:17] We go and prefetch every page on your site. And essentially you're doing your build at runtime. So we're going to try to prevent these cases from happening, but just natively going through DSG everything is not my recommendation. That's for sure. [00:25:32] Not today. At least not today. [00:25:35] swyx: so a couple of things on that. So, this Gatsby image point is very interesting. So how does Gatsby image work with DSG? [00:25:42] Abhi Aiyer: So yeah it works how it does it in Gatsby build. currently today Gatsby uses Gatsby-plugin-sharp and the sharp transformers to take one image, turn it into another. [00:25:54] And even in Gatsby cloud, before we implemented parallel image processing, images were like the slowest part of the Gatsby build because a lot of time, a lot of memory, et cetera. And so we solved that problem. And so in the DSG runtime, we do image processing there for a particular page. [00:26:15] So you will have to wait for image processing. If you're image heavy on a DSG page. [00:26:21] swyx: Which I mean, does that mean that you cannot do a DSG in a serverless function? [00:26:26] Abhi Aiyer: In a total? We do. We actually do DSG in serverless. And that's totally fine. Like you can do image processing, you know? But like, I would say your mileage may vary given what kind of transformations you have going on, how many images you have, right. [00:26:42] But like you said, there's, trade-offs right. If the page makes sense for it, you know, we have a bunch of examples that do have images and they work great, you know? But I don't know if I go full on portfolio with like a, you know, like a masonry thing where like, there's like tons of images and they have sub images and you have to go, like, I'll be like a carousel of images and stuff that may not be good for your. [00:27:06] I don't know, but the choices, the users, that's, what we're trying to get at is like, we're trying to give as many options. We're going to give guidance and like we're having our own opinions, but you, you can choose to listen or not, or, you know, do your own thing and we should try to support you as much as we can. [00:27:25] Automatic DSG [00:27:25] swyx: Yeah, you give me some thought about like, having sort of like a browsers list type of API where you can say like, okay, top 100 most visited pages. No, this is not it. You know what I mean? Like, as a handholding for what should be DSG and what should be statically generated you know, plug into my Google analytics, tell me like top hundred pages statically render those, everything else, DSG. [00:27:48] I'm sure you've thought about it. And I think like maybe four years ago, Gatsby and Guess.js had someone in collaboration, which I assume went nowhere. But let me know if there's. [00:27:59] Ward Peeters: Uh, okay. [00:28:02] For now. Yeah, because there is a new way to do it because now greet guests, it stored everything in one file. So we have to like sometimes download a five megabyte Jason file to make guess.js work. Mondays switching around that you could make, get smarter to say like a guess for this route. You only need the bit of the JSON file. But we never implemented it. So, [00:28:26] Abhi Aiyer: yeah. And we have this, so I'm speaking from the Gatsby cloud perspective, but like you're right, Shawn. Like, if you can hook into Google analytics, you'll get the same thing. [00:28:36] But if you host with Gatsby cloud, we know what, what routes coming through our hosting layer. We know what routes for your site. Are the most hit, you know, we know the requests per route. I mean, how much bandwidth you're using, like per route. And so we could be smarter and tell people exactly how. How to DSG, right? How should you DSG and get it done that way, for sure. [00:29:04] swyx: Okay. So like a, to be, to be complete, uh, typical to be [00:29:08] Abhi Aiyer: complete, you know, we're still in beta forgets before, so I would say like, maybe like after we launched for, for sure, we'll start adding some sugar on. [00:29:17] swyx: Got it. So first of all I did, so this was my first time trying out Gatsby Cloud. I, I think it was behind like a signup wall, like a private beta in the past. And I never really gave it a spin, but again, you know, the V4 announcement really got me going and And yeah. I mean, I'm pretty impressed. [00:29:33] Gatsby Cloud vs Netlify [00:29:33] swyx: So how much of this, you know, the hard question, right? How much of this is a Netlify clone, what are you doing differently? [00:29:40] Abhi Aiyer: Let's talking about that. How much does like DSG is [00:29:45] swyx: how much of Gatsby Cloud? Isn't it [00:29:48] Abhi Aiyer: like? 0%. Ooh, okay. Yeah. Probably 0% of it is a Netlify clone. [00:29:56] swyx: I do like when you provision it, it gives me like a really good set of options. Uh, let's see, uh, you know, connect CMS guests. Netlify does not have that. [00:30:07] Abhi Aiyer: Yeah. I mean, I would, yeah. We're far from an elephant clone Mo multiple weeks. We've built our whole system based on the needs of Gatsby. The way our cloud front end and our back ends talk to our customers, Gatsby Sites is a novel way of doing it. We've exposed that in open source and I think Netlify actually did implement something for external jobs or something with Google pub sub I, I saw that, but everything we do in Gatsby cloud is for Gatsby. We have no other framework that we need to maintain nor care about, sorry. Luke's or whatever. Like we don't care about that. On Gatsby cloud, we've optimized our hosting layer with Fastly to be part of the data. And so if Gatsby changes, Gatsby cloud changes, and if we need to get to be framework to change, it will for Gatsby cloud. So, and we use Google cloud, so we're not on AWS. [00:31:09] I would say we have the similar features though, and that's a valid point to bring out. [00:31:13] We have, we have functions, right. [00:31:15] We have domains and we don't have a purchasing domains or anything yet, but you know, we have the whole hosting product and everything like that. Yeah. [00:31:27] swyx: Is that, is that what you would need for Gatsby Cloud to come out of beta? Like. Domains or like what, what, what are you waiting [00:31:35] Abhi Aiyer: for essentially? Well, Gatsby cloud isn't in beta. [00:31:38] It's like a [00:31:38] Oh Gatsby v4 [00:31:40] swyx: is in beta [00:31:41] Abhi Aiyer: yeah. V4 it's in beta. Yeah. Sorry. Yeah. Yeah, domain like domain registry and all that stuff is more sugar on top that we'll be probably releasing mid next year. But we're just trying to be I mean, Gatsby cloud, from that perspective, we just want to be at the table with all the other JAMstack providers. [00:31:59] But our edge is if you want to build a Gatsby site the best way, you know, you have our support team, that'll help you. Right. As a customer of ours, you're like our family. The Gatsby family, you know, we're, we'll help. We help our customers. We have great support and everything we do on the platform is for Gatsby and making Gatsby better. [00:32:18] So there's like so many things in the framework that we've improved by having Gatsby cloud. Cause we didn't know all the sites that could exist and not do things nicely or have problems or, you know, because of Gatsby cloud that the framework is getting so much better because we have real users feedback and they have a lot of demands and we like to, you know, fulfill them. Yeah. [00:32:41] swyx: Okay. Actually I should probably clarify this earlier. How much of what we just talked about is Gatsby Cloud-only? [00:32:48] Abhi Aiyer: Pretty much all of it, DSG, SSR, they're all capable, you know, you can run it locally and stuff. And I know Netlify has a Gatsby plugin as well that will allow you to run DSG and SSR as well. [00:33:03] For those who are not using those platforms, it's like maybe you're using Amplify or whatever. You're going to have to implement this yourself. I don't recommend it though, because it was a pain in the ass to put it together. But yeah, it should work the best on Gatsby cloud. [00:33:19] Ward Peeters: So technically all of that we building with v4 is in open source. [00:33:22] So you could wire it up all yourself, but why bother if you can use, like Gatsby Cloud. Yeah, you don't have to care about it. [00:33:34] Gatsby vs Next.js [00:33:34] swyx: That's true. Okay. So, just on the rendering side of things, right? I made this observation that, Gatsby now has SSR, now has serverless rendering. All the different rendering modes, like this looks very similar to next JS. [00:33:48] Is it possible to basically say like, okay, Gatsby is the data layer and is this the best data layer and most advanced or whatever, because this is basically what Next.js does, right? Like it's a very, very constrained rendering layer. Why can't you, I mean, you know, sunk costs aside. Why can't you be a next JS layer? [00:34:08] Ward Peeters: Well, technically we could now, because they like implemented everything too, like they have SSG, they have ISR, they have SSR and we could technically move the data layer out of, and use it with next. That could be a possibility, but. We've been, we've come so far and I think do already have built this. [00:34:31] And then now they're also parity. I think having two separate ones and having different dev experience, and maybe Next.js is winning now and Gatsby will a win in, in two months or vice versa. I think it's just a healthy balance. Like it's and I think it's the same thing as a browser wars, like everyone is going to Chrome or chromium and then there is still like, Firefox and iOS, but how long will they survive? [00:34:58] And I think just the competition is what you need. And I think that's why a good reason why we keep separate. And also, I don't think that Next.js is for like, merging with Gatsby or like having the like the same. [00:35:13] swyx: Oh, I think I know Next.js, it be super happy about it, because then they, when they, when the server for reacts you know, role, and then you focus on the data role, right? [00:35:22] Like, uh, Makes sense to me, obviously I'm brushing over a lot of the plugins actually have a rendering portion as well. So how much can you separate those things [00:35:33] Abhi Aiyer: if in the next. No, this is possible. I don't, I mean, we're not going to like say that it's happening or anything. [00:35:41] Gatsby and the Content Mesh [00:35:41] Abhi Aiyer: Like if we look at Gatsby's like, this is how it's set up. [00:35:45] It's, it's what we call the content mesh. You have all these different data warehouses that exist. WordPress Drupal, et cetera, can even be a freaking Google Sheets. You know, like whatever, and we assemble this data layer at build time. And in doing DSG and SSR, we build something called the query engine that allows you to query this LMD B store that has like the manifested data in there. [00:36:13] So. It really opens up the gate for yeah. If you want to use our data layer in a Next.js app, like, I mean, go ahead. Like once we expose this API to our customers then you can essentially have Gatsby data in an iOS app or an Android app react native. Like, it's just an API call at that point. And you know, Gatsby cloud hosts, like a graphical API for you that you can just query your data. [00:36:38] I don't know if any data scientists would care for that. They could add that into Looker or something. You know, like I remember they want to do it like that stuff would be available and it's almost like a content data lake versus, you know, traditional data lake I guess. It's purely for content and you would have the benefits of Gatsby because we normalize and we create structures and you like, the user can customize a schema, however you want. [00:37:05] And then now you can use it on multiple platforms, right? It's not an immediate goal for us to do so. It's a logical next step. Yeah. Yeah. [00:37:15] swyx: Awesome. Awesome. Cool. Yeah, I, I feel like that's a really good and in depth coverage. [00:37:19] React 18 and Gatsby [00:37:19] swyx: Maybe let's end off with talking about the future of React 18 and your plans there. First of all, what's happening in react 18. Is it out? Like the plan for the react 18 and published in June? Okay. All right. Let's talk about it. What's what's going on? [00:37:35] Ward Peeters: So, yeah, so we are working closely with the React team and we also in the working group to figure out like, okay, how can we help the team, make it more stable and give it in user hands. [00:37:46] So I think from may or something, we have introduced React 18 as part of Gatsby. So you can now install React 18 alpha. And we just moved to the new rendering mode. So the async mode suspense and all those things were. Like what, what we're planning on, at least when you use Gatsby, like we have page queries and we have static queries and there's a big pain point is static queries, cause it's a graph QL query, but you cannot have any variables, which means you're kind of limited to the unit. And then you have to move everything to page queries going to have to know all the content up front and wait the new async rendering bits of React to get into like a useQuery, because you can yield the rendering of React at any time. [00:38:34] Cause async doesn't mean you have to go like, uh, use Apollo Server to get server data tree or something or other pieces, or you kind of have two have React async mode or React Suspense in SSR and we can all move it to the page components or the components of your reactor. So basically look that you're just recreating an react application and then every async bit like using react-fetch or a useQuery, it all just works. [00:39:02] I think that's where, where we activate in benefits a lot where it's. It just removes a lot of cruft or that you have to do now. It gets you where you have to be in the Gatsby mindset when you're developing and, and you basically go to a, creating a react app and you have a data layer, but I think React 18 opens so many doors with the new cache APIs. It just becomes way smarter and when you look at it from a performance perspective with the whole concurrent mode where inputs gets priority over rendering, it's just going to be way smoother than what they had so far. [00:39:39] Abhi Aiyer: And hopefully people stop complaining about lighthouse scores and stuff. That'll be great. [00:39:45] Custom rendering page fragments with React 18 [00:39:45] Abhi Aiyer: Another cool thing that React 18 kind of unlocked for Gatsby in particular is a concept of fragments. And so we were talking about that nav bar example earlier with the a hundred thousand pages. And we want to leverage react 18 with like custom renderers so that we can essentially create fragments of a page that had beta dependent. [00:40:07] Because there's no page query or static query anymore. That's just a query. Your navbar component has a query and essentially Gatsby can make that nap bar a navbar fragment and your body has a fragment, or maybe your footer has a fragment. Your sidebar has a fragment. And as data changes incrementally, we only rebuild fragments and our hosting layer, stitches, fragments together. This is an old concept called ESI includes like if everyone did PHP back in the day, like, you know, very familiar with this stuff, like I said, every 10 years, things has come back around and we're going to try to do that. We're going to try to build fragments of pages, stitch them together. So a navbar change doesn't break the bank, you know? But we can only do that once react 18. It's like, you know, fully there. I mean, we could do it now, but like why, when we should just like work off the, the, the work of others. [00:41:02] swyx: So when you say fragments, are you referring to GraphQL Fragments or, or like [00:41:06] Abhi Aiyer: Asian fragment might be a, maybe we call it like, you know, today, like an HTML page that has specific. [00:41:13] You know, I like to call him like the rectangles that we all draw around are our websites. Right. They all have independent data isolation. Right. And so these are like what maybe a Gatsby slice of a page or a fragment or some type of include, you know, like in the templating days. Right. And that's what I kind of mean there. [00:41:31] So these includes or templates or whatever you want to call them would be independently built. And then independently stitched at the cache layer. And then, you know, the data dependencies don't cross, and now I'm not building a hundred thousand pages because I misspelled Gasby and it should've been, you know, [00:41:51] swyx: sounds like it happens a lot, [00:41:54] Abhi Aiyer: but definitely those, [00:41:56] Ward Peeters: and it looks a lot like donut caching. [00:41:58] If you're more familiar with that piece, like you have a page where I said parks has a different. Limit and another one. So that's more or less the technical piece out of [00:42:10] Server Components in Limbo [00:42:10] swyx: a server components. Anything on any implications on that forgets me? [00:42:15] Ward Peeters: Not yet. I would say because they're not going to ship it with react 18. [00:42:19] We've been talking about it, but it's still very fresh or very new, like even the React team hasn't, hasn't worked more on it, so they did their demo, but then it's got like a little bit [00:42:31] swyx: stagnated. Oh my God. [00:42:37] Ward Peeters: All the pieces. Like they need to build all the pieces underneath it to make it work. [00:42:45] swyx: They jumped, they jumped the gun, maybe in announcing I got so excited. I was like, wow. Okay. I can cut my Javascript bundle by 40% and run backend functions in my react component. And then nothing, nothing for 10 months, [00:43:01] Ward Peeters: because we are super excited about it too. Because when you look at especially marketing sites, like marketing pages or blogs, there's only a small piece of JavaScript that you actually need. [00:43:13] Like maybe you need a bit for your newsletter button or you like something like that. And why. 200 kilobytes of JavaScript could bring technically only need maybe 10, 20 kilobytes. So I think it's static or with like marketing pages. Uh, [00:43:33] Smart Servers vs Smart Clients [00:43:33] Abhi Aiyer: yeah, so the world was server rendered. Then we went client side rendered. Then we went static rendered. Now we're DSG rendered, and then we're going to go back to server run. So, you know, time just keeps spinning. Partially server. [00:43:47] swyx: I called it smart server versus smart clients is my term for it. So this is the, I think maybe my, my most recent posts, because I have been trying to write more, but then I keep have having real life get in the way. [00:44:01] But why is traditional, which is server rendered, different from the new server rendered. We have essentially is essentially exactly the same, but there's a thin runtime, which I'll ship the stuff that we send over the wires changes. And we actually doing rendering in the browser, but like partial rendering, maybe I should say. [00:44:20] And yeah. I dunno. I think, I think this is a very interesting exploration. Phoenix live view is also the other one that, that gets a lot of love for this. And then rails is also adopting Hotwire. So, I don't know where this goes. I mean, I, I it's, it seems like we fully explored the smart client space and the smart server revolution is just kind of get, getting going. [00:44:41] Ward Peeters: We're going back to Meteor. [00:44:44] swyx: Back to meteor, but not so opinionated, I think, you know, I was very excited about meteor. Like when I, when I first started as a web dev, I was like, oh yeah. Okay. Everything is in there. I actually mentioned Meteor here because it had the mini Mongo data store, which was I thought it was just such a great experience. [00:44:59] Did you use. [00:45:02] Abhi Aiyer: Oh, both my last company, we used meteor for our backend, and then we had to kind of migrate slowly off of it. Cause they were just ahead of their time. You know, now all those concepts. Those are like, those are the concepts of today. Right. And that's the beautiful thing they were [00:45:19] swyx: just ahead of their time. [00:45:21] Apollo and Open Source Startup Strategy [00:45:21] swyx: I mean, you know, what they did was they became Apollo. They were just like, oh no, one's no, one's handling all the hard parts of GraphQL. Well, [00:45:29] Abhi Aiyer: okay. We'll do it. Yeah, good job of that too, [00:45:33] swyx: which is by the way, like in terms of just honestly, I'm interested in startups, entrepreneurship, uh, you know, we worked so hard in web dev stuff. [00:45:41] A lot of this, we never charge a cent for and something I would like to make money on the smart things that we do in tech. [00:45:47] Taking an under specified spec, which most of the times is intentionally under specified, and then building all the hard parts around it, is a very interesting formula for success. [00:45:58] So essentially React and under specified framework and Next.js came in and went like, oh, okay, well, we'll build the get initial props that you guys forgot. And great, very successful Gatsby, same thing. And then Apollo and Relay by the way, but, but relay was not a serious company, a company effort. [00:46:19] I mean, Relay is a serious effort. It's not a startup that was like existentially relying on like, uh unsuccess. Whereas was Apollo was like, okay, GraphQL was under specified. There's a reference JS implementation, but no one's building the production quality standard. We'll do it. And then, and yeah, like it's really interesting. Cause as the spec grows or as adoption of the thing grows, you're you grow with it and, you serve the audience and you also capture a lot of the value and you essentially have Facebook working for you in the sense of like, oh, there's the spec maintainers, you know, whatever, whatever the spec is, they're working for you because every time they contribute to the spec, you. [00:47:06] TMA: Too Many Acronyms [00:47:06] Abhi Aiyer: Yeah, maybe that's what the what's going to happen with DPR. Right? [00:47:10] swyx: The naming socks, too many, three letter acronyms. I'm sure. Like, look like you and I, and everyone in like the WebDev, like Twitter sphere or whatever, we don't mind new things and like understanding the differences in nuances, but anyone who is like just a regular web dev or just like not web dev, but talking to web devs, they think we're crazy. [00:47:36] This is actually bad. Like it, we look like the nerds, uh, who. Talking about all these minor differences and inventing new acronyms for them. I don't know how to fix it. Jargon is important for specialists to understand in a very short amount of time, the differences between what we referring to. Jargon is important, but we don't do ourselves, our industry a favor when we have all these acronyms and then people just throw them on onto a page or a blogpost or a slide deck. [00:48:05] And then. People would just go like, okay. Yeah, the JS ecosystem [00:48:09] Abhi Aiyer: is crazy. And you ended up explaining the same thing all the time. Right? Cause you use some acronym. It was funny, like on the way to Gatsby camp, like we had, like all of our release had all of the releases and gas before had the acronym. Yeah, like PQR parallel query, running DSE, SSR, SSG, man. [00:48:26] We were like trying to figure it out. How many more acronyms can we fit to, to get like the, the acronym count up, but it's a serious problem for us too, because our, some of our customers have never used Gatsby before they're coming from a WordPress full on WordPress background and our sales team marketing, we all need to be able to convey like, yeah, this is what it really is. [00:48:45] And this is what it means. And maybe. The acronym sticks after they understand it, but that's a really uphill battle to explain right on the way. So I would love if a community we all got together and like, kind of just understood it. You know, it's kind of like the GraphQL spec have a formal definition for what this is. [00:49:02] Don't be too heavy handed on approach, let people implement however they want to. And then there's just a concept that has different flavors. Yeah. Oh, it's different [00:49:14] swyx: flavors. Okay. That'd be interesting. [00:49:16] Gatsby for Docs [00:49:16] swyx: Is there anything else that we haven't covered that you wanted to shout out? [00:49:21] Abhi Aiyer: This is fun. I really enjoyed talking to you too. [00:49:24] swyx: Yeah, I love, uh, I love catching up. Um, uh, Fun fact, we're actually at my workplace. We use Docusaurus right now for our docs. We're actually considering moving to Gatsby. [00:49:35] Nice. Not something I thought I would do this year, but we're, we're running into enough limitations to Docusaurus that we're essentially customizing so much that we don't get much benefit anymore. So maybe a good standard docs implementation. It would be interesting for you guys actually, because a lot of the reason that people pick Docusaurus is basically it has docs in the name and it's got a lot of good defaults for docs, right? [00:50:04] And Gatsby, maybe it doesn't have such a developed theme for docs. [00:50:07] Ward Peeters: We've mostly pushed people to the Apollo team. Like they have a great, like the whole Apolo site is, or docs site is built with Gatsby and a open source. The building blocks up there. So, or you could start from there and then, oh [00:50:20] Abhi Aiyer: yeah. [00:50:23] New Relic is with Gatsby and they're working on something similar too. [00:50:30] swyx: Awesome. Awesome. Yeah. All right. Cool. Well thanks for those pointers. I'm actually going to go explore them. [00:50:38] Abhi Aiyer: Yeah. If you need any help. Yeah, we'll do. [00:50:41] swyx: And there's no reason why we shouldn't move to Gatsby cloud, if that makes sense for us as well. Okay. Okay. [00:50:47] Ward and Abhi,thanks so much, and this is really great chatting, thanks for reaching out. And, yeah, I hope [00:50:52] Abhi Aiyer: people would try out Gatsby. [00:50:54] Thanks for having us.
im Dilettoso is one of the best known figures in UFO and Exotic Science Research. For over 30 years he has been recognized as one of the leading analysts of UFO Pictures and Video. He has tested over 2000 photographs and 400 video cases. He has appeared on many TV Broadcasts, including Sightings, Discovery, A&E, Disney, MTV, CNN, UFO Hunters and the BBC.He is known in the UFO community as an innovative researcher and media communicator. Never without controversy, his work on the Billy Meier case, the Phoenix Lights, and Mexico City UFOs have raised the visibility of UFOs into the mainstream. He is the co-founder of the International UFO Congress In his other life, at Village Labs, Jim is a specialist in Image Processing and Bio-Bio Acoustics. He has worked with NASA, Allied-Signal and the Moody Blues. Currently he is engaged in the technology of agriculture and the electro-acoustic stimulation of plants.His email is jimtoso777@gmail.comhttps://www.linkedin.com/in/jim-dilettoso-19b60310/
We return to book reviews in this episode and discuss and review Ragnar Axelsson's masterpiece: Faces of the North. Faces of the North is an outstanding example of a documentary photography book and rates five stars. Faces of the North is currently out of print; however, I highly recommend you consider adding it to your photographic library if you can find a copy.Faces of the NorthI have also added two new levels of support for the Podcast that include monthly image critiques for those of you who would like feedback on your work and one-on-one one hour mentoring sessions on Post production and Image Processing.Support the show (https://www.buymeacoffee.com/JoshuaHolko)
In this podcast episode, we discuss how to create great wildlife photographs in the field. We also touch on several earlier podcasts including:The Three F's of Wildlife PhotographyWhy a new camera might be detrimental to your photographyI have also added two new levels of support for the Podcast that include monthly image critiques for those of you who would like feedback on your work and one-on-one one-hour mentoring sessions on Post-production and Image Processing.Support the show (https://www.buymeacoffee.com/JoshuaHolko)
In this episode, we discuss the COVID-19 situation in Australia as of mid-August 2021 and what it means for travel for Australian citizens. We also discuss the recent feature portfolio published in the Hungarian Digitalis Photography magazine as well the feature article published by UK Business Influencer magazine on the 2022 Winter expedition I am leading with my good friend David Gibbon to Ellesmere Island in search of the elusive White Arctic Wolf. Lastly, we also discuss an invitation to be the guest speaker and open an exclusive new gallery here in Melbourne later this year.UK Business Influencer Magazine feature article on 2022 Ellesmere Island Expedition Hungarian Photography magazine Portfolio Feature - Ice Cold PassionI have also added two new levels of support for the Podcast that include monthly image critiques for those of you who would like feedback on your work and one-on-one one-hour mentoring sessions on Post-production and Image Processing.Support the show (https://www.buymeacoffee.com/JoshuaHolko)
In this episode, we return to book reviews and discuss and review Vincent Munier's wonderful (but sadly out of print) Kamchatka. For those of you who may be unfamiliar with Kamchatka; it is located in a remote part of far eastern Russia, lying between the Sea of Okhotsk on the west and the Pacific Ocean and the Bering Sea on the east. It's renowned for its supervolcano, dramatic landscapes, and wild bears. If you can find a copy of Kamchatka in a second-hand book store anywhere I highly recommend you grab it at pretty much any price. Book Review Born to Ice by Paul NicklenBook Review Night of the Deer by Vincent MunierBook Review Arctic by Vincent MunierHow to Take Better PhotographsI have also added two new levels of support for the Podcast that include monthly image critiques for those of you who would like feedback on your work and one-on-one one hour mentoring sessions on Post production and Image Processing.Support the show (https://www.buymeacoffee.com/JoshuaHolko)
Sorularınız için: https://bit.ly/3xuKeBR Canlı yayın linki: https://bit.ly/3dYvFhZ Kariyer sohbetlerini takip etmek için: https://bit.ly/2HuqQya Websitemiz: https://kesisenyollar.org/ Youtube kanalımızı takip etmek için: http://bit.ly/KesisenYollarYoutube Geri bildirimleriniz için: https://bit.ly/3xvWe69 Gözde Ünal Lisans derecesini ODTÜ, Yüksek Lisans derecesini Bilkent Üniversitesi'nden aldıktan sonra doktora çalışmalarını 2002 senesinde North Carolina State University'de Elektrik ve Bilgisayar Mühendisliği bölümünde, Matematik doktora yandalı ile birlikte aldı. Ardından Georgia Institute of Technology'ye doktora sonrası araştırmacı olarak katıldı. Daha sonra 2003-2007 seneleri arasında Siemens Corporate Technology, Princeton, New Jersey'de araştırmacı bilim insanı olarak çalıştı.2007 senesinde Türkiye'ye akademisyen olarak döndü. İlk olarak Sabancı Üniversitesi'nde Mühendislik ve Doğa Bilimleri Fakültesi'nde Yrd. Doç. Dr. ve Doç. Dr. olarak çalıştı. Ardından İstanbul Teknik Üniversitesi Bilgisayar ve Bilişim Fakültesi'ne katıldı, ve halen bu kurumda Profesör olarak görev yapmakta. 2018-2020 arasında İTÜ Yapay Zeka ve Veri Bilimi Uygulamalı Araştırma Merkezi'nin (ITU-AI) Kurucu Direktörü olarak görev yapan Prof. Ünal uluslararası ve ulusal birçok organizasyonda görev aldı. Bunlar arasında MICCAI 2016, MIDL 2019 ve SİU 2017 konferansları Teknik Program Başkanlığı, Women in MICCAI kurucu üyeliği, IEEE Transactions on Image Processing dergisinde Yardımcı Editörlükleri sayılabilir. 30'dan fazla uluslararası patente sahip olan Prof. Ünal, 2010 senesinde L'Oreal Türkiye Yılın Genç Bilim Kadını ve TÜBA GEBİP Genç Bilim İnsanı ödülleri, 2016 senesinde Marie Curie Alumni Association Kariyer Ödülünü almıştır. 2020 senesinde öğretime başlayan İTÜ Yapay Zeka ve Veri Mühendisliği Lisans Bölümü kurucu profesörlerindendir. Halen ITU AI kurucu üyesi ve ITU Vision Lab direktörlüğünü yürütmekte olan Prof. Ünal yapay zeka, derin öğrenme ve bilgisayarla görü alanlarında araştırma çalışmalarına devam etmekte.
Przemek Matylla talks about building an image processing API with C, Python and Node. It's hosted on bare metal servers.
- GM Accused of Not Doing Enough Business with Black Media - Chinese EV Startups Have a Good Day - Traditional Automakers Face Big Risk Switching to Electric - Cadillac Getting Electric D-Sized Crossover - Hong Guang MINI EV Gets a New Variant - BMW Helps Improve Image Processing - Daimler Lays Claim to First Pickup Truck - Ranking Traditional OEMs by U.S. BEV Market Share - Mazda MX-5 Miata Drive Review - Chrysler Pacifica Stow N' Go Correction - Do You Know What Car This Is?
- GM Accused of Not Doing Enough Business with Black Media- Chinese EV Startups Have a Good Day- Traditional Automakers Face Big Risk Switching to Electric- Cadillac Getting Electric D-Sized Crossover- Hong Guang MINI EV Gets a New Variant- BMW Helps Improve Image Processing- Daimler Lays Claim to First Pickup Truck- Ranking Traditional OEMs by U.S. BEV Market Share- Mazda MX-5 Miata Drive Review- Chrysler Pacifica Stow N' Go Correction- Do You Know What Car This Is?
Backpropagated Gradient Representations for Anomaly Detection, Implicit Saliency in Deep Neural Networks, Contrastive Explanations in Neural Networks, Fabric Surface Characterization: Assessment of Deep Learning-Based Texture Representations Using a Challenging Dataset, Successful Leveraging of Image Processing and Machine Learning in Seismic Structural Interpretation, and Relative Afferent Pupillary Defect Screening Through Transfer Learning. Prof. Ghassan AlRegib is Professor in the School of Electrical and Computer Engineering at the Georgia Institute of Technology. He is the director of the Omni Lab for Intelligent Visual Engineering and Science and the center for Energy and Geo Processing at Georgia Tech. --- Send in a voice message: https://anchor.fm/scientificsense/message Support this podcast: https://anchor.fm/scientificsense/support
Dr. Enrique Saldívar received his medical training in Universidad La Salle, received his MD from Universidad Autonoma de Mexico (UNAM), his Masters in Biomedical Engineering from Universidad Autonoma Metropolitana, and his Ph.D. in Bioengineering from the University of California, San Diego. Throughout his career, he has been appointed faculty at The Scripps Research Institute, La Jolla Bioengineering Institute, The West Wireless Health Institute, and Case Western Reserve University. His expertise includes: Biomechanics, Microcirculation, Rheology, Platelet Engineering, Digital Signal Processing, Image Processing, Bio-Micro-Electro-Mechanical Systems (Bio-MEMS), Nanotechnology, and Digital Health. Dr. Saldívar career in medical devices has focused on the transformation of cutting-edge technological developments into meaningful medical solutions. His multidisciplinary background combined with a deep sense of social responsibility provides him with a unique perspective to provide solutions to unmet medical needs in underserved communities. Dr. Saldivar’s interests are focused on the use of technology to improve the quality of life, globally, and to ameliorate the accessibility to first-class medical attention in underprivileged communities. Dr. Saldívar scientific career has focused on the study of biomechanical mechanisms responsible for complex physiological responses with an emphasis on the rheological mechanisms, at both the cellular level and at the cell membrane level. He has made seminal contributions to the understanding of platelet adhesion under flow and chronic adaptation to extreme hypoxia.
Sponsored by us! Support our work through: Our courses at Talk Python Training pytest book Patreon Supporters Special guest: Jason McDonald Michael #1: 5 ways I use code as an astrophysicist Video by Dr. Becky (i.e. Dr Becky Smethurst, an astrophysicist at the University of Oxford) She has a great YouTube channel to check out. #1: Image Processing (of galaxies from telescopes) Noise removal #2: Data analysis Image features (brightness, etc) One example: 600k “rows” of galaxy properties #3: Model fitting e.g. linear fit (visually as well through jupyter) e.g. Galaxies and their black holes grow in mass together Color of galaxies & relative star formation #4: Data visualization #5: Simulations Beautiful example of galaxies colliding Star meets black hole Brian #2: A Visual Intro to NumPy and Data Representation Jay Alammar I’ve started using numpy more frequently in my own work. Problem: I think of np.array like a Python list. But that’s not right. This visualization guide helped me think of them differently. Covers: arrays creating arrays (I didn’t know about np.ones(), np.zeros(), or np.random.random(), so cool) array arithmetic indexing and slicing aggregation with min, max, sum, mean, prod, etc. matrices : 2D arrays matrix arithmetic dot product (with visuals, it takes seconds to understand) matrix indexing and slicing matrix aggregation (both all entries and column or row with axis parameter) transposing and reshaping ndarray: n-dimensional arrays transforming mathematical formulas to numpy syntax data representation All with excellent drawings to help visualize the concept. Jason #3: Qt 6 release (including PySide2) Qt 6.0 released on December 8: https://www.qt.io/blog/qt-6.0-released 3D Graphics abstraction layer called RHI (Rendering Hardware Interface), eliminating hard dependency on OpenGL, and adding support for DirectX, Vulkan, and Metal. Uses native 3D graphics on each device by default. Property bindings: https://www.qt.io/blog/property-bindings-in-qt-6 A bunch of refactoring to improve performance. QtQuick styling CAUTION: Many Qt 5 add-ons not yet supported!! They plan to support by 6.2 (end of September 2021). Pay attention to your 5.15 deprecation warnings; those things have now been removed in 6.0. PySide6/Shiboken6 released December 10: https://www.qt.io/blog/qt-for-python-6-released New minimum version is Python 3.6, supported up to 3.9. Uses properties instead of (icky) getters/setters now. (Combine with snake_case support from 5.15.2) from __feature__ import snake_case, true_property PyQt6 also just released, if you prefer Riverbank’s flavor. (I prefer official.) Michael #4: Is your GC hyper active? Tame it! Let’s think about gc.get_threshold(). Returns (700, 10, 10) by default. That’s read roughly as: For every net 700 allocations of a collection type, a gen 0 collection runs For every gen 0 collection run, 1/10 times it’ll be upgraded to gen 1. For every gen 1 collection run, 1/10 times it’ll be upgraded to gen 2. Aka for every 100 gen 0 it’s upgraded to gen 2. Now consider this: query = PageView.objects(created__gte=yesterday).all() data = list(query) # len(data) = 1,500 That’s multiple GC runs. We’ve allocated at least 1,500 custom objects. Yet never ever will any be garbage. But we can adjust this. Observe with gc.set_debug(gc.DEBUG_STATS) and consider this ONCE at startup: # Clean up what might be garbage gc.collect(2) # Exclude current items from future GC. gc.freeze() allocs, gen1, gen2 = gc.get_threshold() allocs = 50_000 # Start the GC sequence every 10K not 700 class allocations. gc.set_threshold(allocs, gen1, gen2) print(f"GC threshold set to: {allocs:,}, {gen1}, {gen2}.") May be better, may be worse. But our pytest integration tests over at Talk Python Training run 10-12% faster and are a decent stand in for overall speed perf. Our sitemap was doing 77 GCs for a single page view (77!), now it’s 1-2. Brian #5: Top 10 Python libraries of 2020 tryolabs criteria They were launched or popularized in 2020. They are well maintained and have been since their launch date. They are outright cool, and you should check them out. General interest: Typer : FastAPI for CLI applications I can’t believe first commit was right before 2020. Just about a year after the introduction of FastAPI, if you can believe it. Sebastián Ramírez is on
Nick Chase is the author of the video series “Machine Learning for Mere Mortals.” He helps break down some of the misconceptions about how complicated Machine Learning is and the magical parts of the science. He and the panelists then dive into the basics of what you need to know and break up the scary sounding terms and mathematical concepts into bite size pieces. Enter for a chance to win a copy of Nick’s video course at https://devchat.tv/mere-mortals. Sponsors Machine Learning for Software Engineers by Educative.io Audible.com CacheFly Panel Charles Max Wood Jason Mayes Guest Nick Chase Picks Jason Mayes: Browser-Based Augmented Reality Sudoku Solver using TensorFlow and Image Processing Charles Max Wood: https://www.brandonsanderson.com/the-stormlight-archive-series/ https://developeronfire.com/podcast/episode-448-nicholas-chase-expanding-what-people-think-they-can-do Nick Chase: Machine Learning for Mere Mortals Follow Adventures in Machine Learning on Twitter > @podcast_ml
In the fourth episode of the "You Belong in AI!" podcast, UCLA ACM AI Outreach interviews Shweta Khushu (she/her/hers). Shweta is a Sr. Engineer and Growth Lead of the AI team at SkySpecs, a startup based in Ann Arbor, MI that uses AI and robotics to fully automate drone inspections of wind turbines to help the green energy initiative so that one day renewable energy will be ubiquitous. She received her MS degree in Electrical and Computer Engineering from the University of Michigan in 2016 with a specialization in Machine Learning and Image Processing, despite beginning her journey in STEM in Electronics Engineering. Shweta is passionate about solving real-world problems in the Computer Vision space. Outside of work, she trains in Indian classical music, and likes to bake, read, watch movies, and spend time with her tabby cat Oliver. In this episode, Shweta discusses the importance of representation and inclusion in tech and AI. Additionally, she provides advice and words of encouragement to youth from underrepresented groups who are interested in AI but not sure how to get started. The "You Belong in AI!" podcast is made possible by UCLA ACM AI Outreach. All questions are contributed by Arjun Subramonian (Outreach Director 20-21), Maya Raman (Events Director 20-21), Kai Tota (Outreach/Events Officer 20-21), Jason Jewik (Outreach Officer 20-21), Mat Ruíz (Outreach/Events Officer 20-21), Aman Oberoi (Outreach Officer 20-21), and Nisha McNealis (Outreach Officer 20-21). The podcast is edited by Arjun Subramonian. The music you hear in this episode is: Cheery Monday by Kevin MacLeod Link: https://incompetech.filmmusic.io/song/3495-cheery-monday License: http://creativecommons.org/licenses/by/4.0/ Want to learn more about UCLA ACM AI Outreach? Visit our website. Want to learn more about Shweta? Make your way to her LinkedIn and/or Twitter. Curious about all the amazing things SkySpecs does? Visit their website.
Todd and Dalton discuss their high level procedures on what they do once underwater photos come out of the camera and into the computer. Backup strategy, photo editing software choices, intelligent folder naming standards are discussed along shot selection and editing techniques.Check out more detailed show notes on our blog.Feedback:If you have thoughts, ideas or comments, email us at feedback@theaquaticlifepodcast.comHelp Us Grow The Community:*** Please take the time to rate and review us on Apple Podcasts or wherever you get your podcast content from. It would help us to grow the community and means a lot to us. Thanks!***You can subscribe to The Aquatic Life on iTunes, Google Podcasts, Overcast, Spotify, and all major podcast apps (RSS)The Aquatic Life Social Media Pages:Website: TheAquaticLifePodcast.comTwitter: AquaticLifePodFacebook: TheAquaticLifePodcastInstagram: TheAquaticLifePodcastMore About your Hosts:Dalton HammPortfolio: daltonhamm.comDive Instruction: piratedivecrew.comFacebook: daltonhammphotographyInstagram: daltonhammphotographyTodd ReimerPortfolio: toddreimerphoto.comFacebook: todd.s.reimerInstagram: tsreimerTwitter: tsreimer
Is AV1 more efficient than HEVC? Dror & Mark get into the middle of a 3 against 1 standoff over whether AV1 is actually more efficient than HEVC. The following blog post first appeared on the Beamr blog at: https://blog.beamr.com/2018/11/23/codec-efficiency-is-in-the-eye-of-the-measurer-podcast/ When it comes to comparing video codecs, it's easy to get caught up in the “codec war” mentality. If analyzing and purchasing codecs was as easy as comparing fuel economy in cars, it would undoubtedly take a lot of friction out of codec comparison, but the reality is that it's not that simple. In Episode 02, The Video Insiders go head-to-head comparing two of the leading codecs in a three against one standoff over whether AV1 is more efficient than HEVC. So, which is more efficient? Listen in to this week's episode, “Codec Efficiency Is in the Eye of the Measurer,” to find out. Want to join the conversation? Reach out to TheVideoInsiders@beamr.com. TRANSCRIPTION (lightly edited to improve readability only) Mark Donnigan: 00:41 Hi everyone I am Mark Donnigan and I want to welcome you to episode two of the Video Insiders. Dror Gill: 00:48 And I am Dror Gill. Hi there. Mark Donnigan: 00:50 In every episode of the Video Insiders we bring the latest inside information about what's happening in the video technology industry from encoding, to packaging, to delivery, and playback, and even the business behind the video business. Every aspect of the video industry is covered in detail on the Video Insiders podcast. Dror Gill: 01:11 Oh yeah, we usually do cover everything from pixels, to blocks, to microblocks, to frames, to sequences. We go all the way up and down the video delivery chain and highlight the most important things you should know before you send any video bits over the wire. Mark Donnigan: 01:28 In our first episode we talked about a very hot topic which asked, “Hasn't this kind of been worn out?” The whole HEVC, AV1 discussion. But I think it was very interesting. I sure enjoyed the talk. What about you Dror? Dror Gill: 01:47 Yeah, yeah, yeah. I sure did. It was great talking about the two leading codecs. I don't want to say the word, codec war. Mark Donnigan: 01:58 No, no, we don't believe in codec wars. Dror Gill: 01:59 We believe in codec peace. Mark Donnigan: 02:00 Yeah, that's true. Why is it so complicated to compare video codecs? Why can't it be as simple as fuel economy of cars, this one gets 20 miles per gallon and that one gets 30 and then I make a decision based on that. Dror Gill: 02:15 I wish it was that simple with video codecs. In video compression you have so many parameters to consider. You have the encoding tools, tools are grouped into what's called profiles and levels, or as AV1 calls them “experiments.” Mark Donnigan: 02:31 Experiments, mm-hmm… Dror Gill: 02:35 When you compare the codecs which profiles and levels do you use. What rate control method? Which specific parameters do you set for each codec? And each codec can have hundreds, and hundreds of parameters. Then there is the question of implementation. Which software implementation of the codec do you use? Some implementations are reference implementations that are used for research, and others are highly performance optimized commercial implementations. Which one do you select for the test? And then, which operating system, what hardware do you run on, and obviously what test content? Because encoding two people talking, or encoding an action scene for a movie, is completely different. Dror Gill: 03:13 Finally, when you come to evaluate your video, what quality measure do you use? There're various objective quality measures and some people use actual human viewers and they assesses subjective quality of the video. On that front also, there're many possibilities that you need to choose from. Mark Donnigan: 03:32 Yeah, so many questions and no wonder the answers are not so clear. I was quite surprised when I recently read three different technical articles published at IBC actually, effectively comparing AV1 versus HEVC and I can assume that each of the authors did their research independently. What was surprising was they came to the exact same conclusion, AV1 has the same compression efficiency as HEVC. This is surprising because some other studies and one in particular (I think we'll talk about) out there says the contrary. So can you explain what this means exactly, Dror. Dror Gill: 04:16 By saying that they have the same compression efficiency, this means that they can reach the same quality at the same bitrate or the other way round. You need the same bitrate to reach that same quality. If you need for example, two and a half megabits per second to encode an HD video file using HEVC at a certain quality, then with AV1 you would need roughly the same bitrate to reach that same quality and this means that AV1 and HEVC provide the same compression level. In other words, this means that AV1 does not have any technical advantage over HEVC because it has the same compression efficiency. Of course that's if we put aside all the loyalty issues but we discussed that last time. Right? Mark Donnigan: 04:56 That's right. The guys who wrote the three papers that I'm referencing are really top experts in the field. It's not seminar work done by a student, not to downplay those papers, but the point is these are professionals. One was written by the BBC in cooperation with the Multimedia and Vision Group at the Queen Mary University of London. I think nobody is going to say that the BBC doesn't know a thing or two about video. The second was written by Ateme, and the third by Harmonic, leading vendors. Mark Donnigan: 05:29 I actually pulled out a couple of phrases from each that I'd like to quote. First the BBC and Queen Mary University, here is a conclusion that they wrote, “The results obtained show in general a similar performance between AV1 and the reference HEVC both objectively and subjectively.” Which is interesting because they did take the time to both do the visual assessment as well as use a quality measure. Mark Donnigan: 06:01 Ateme said, “Results demonstrate AV1 to have equivalent performance to HEVC in terms of both objective and subjective video quality test results.” Dror Gill: 06:10 Yeah, very similar. Mark Donnigan: 06:16 And then here is what Harmonic said, “The findings are that AV1 is not more advantageous today than HEVC on the compression side and much more complex to encode than HEVC.” What do you make of this? Dror Gill: 06:32 I don't know. It sounds pretty bad to me, even two of those papers also analyzed subjective quality so they used actual human viewers to check out the quality. But Mark what if I told you that researchers from the University of Klagenfurt in Austria together with Bitmovin published a paper which showed completely different results. What would you say about that? Mark Donnigan: 06:57 Tell me more. Dror Gill: 06:58 Last month in Athens I was the ICIP conference that's the IEEE International Conference on Image Compression and Image Processing. There was this paper presented by this University in Austria with Bitmovin and their conclusion was, let me quote, “When using weighted PSNR, AV1 performs consistently better for bit rate compared to AVC, HEVC, and VP9.” So they claim AV1 is better than three codecs but specifically it's better than HEVC. And then they have a table in their article that compares AV1 to HEVC for six different video clips. The table shows that with AV1 you get up to 25% lower bitrate at the same quality than HEVC. Dror Gill: 07:43 I was sitting there in Athens last month when they presented this and I was shocked. Mark Donnigan: 07:50 What are the chances that three independent papers are wrong and only this paper got it right? And by the way, the point here is not three against one because presumably there're some other papers. I'm guessing other research floating around that might side with Bitmovin. The point is that three companies who no one is going to say that any of them are not experts and not highly qualified to do a video assessment, came up with such a different result. Tell us what you think is going on here? Dror Gill: 08:28 I was thinking the same thing. How can that be. During the presentation I asked one of the authors who presented the paper a few questions and it turned out that they made some very questionable decisions in all of that sea of possibility that I talked about before. Decisions related to coding tools, codec parameters, and quality measures. Dror Gill: 08:51 First of all, in this paper they didn't show any results of subjective viewing. Only the objective metrics. Now we all know that you should always your eyes, right? Mark Donnigan: 09:03 That's right. Dror Gill: 09:04 Objective metrics, nice numbers, but obviously you need to view the video because that's how the actual viewers are going to assess the (video) quality. The second thing is that they only used the single objective metric and this was PSNR. PSNR, it stands for peak signal-to-noise ratio and basically this measure is a weighted average of the difference in peaks between pixel values of the two images. Dror Gill: 09:30 Now, we're Video Insiders, but even if you're not an insider you know that PSNR is not a very good quality measure because it does not correlate very well with human vision. This is the measure that they choose to look at but what was most surprising is that there is a flag in the HEVC open source encoder which they used that if chosen, the result is improved PNSR. What it does, it turns off some psycho-visual optimizations which make the video look better but reduce the PSNR, and that's turned on by default. So you would expect that they're measuring PSNR they would turn that flag on so you would get higher PSNR. Well, they didn't. They didn't turn the flag on! Mark Donnigan: 10:13 Amazing. Dror Gill: 10:17 Finally, even then AV1 is much slower than HEVC, and they also reported in this data that it was much, much slower than HEVC but still they did not use the slowest encoding standing of HEVC, which would provide the best quality. There's always a trade off between performance and quality. The more tools you employ the better quality you can squeeze out of the video, of course that takes you more CPU cycles but they used for HEVC, the third slowest setting which means this is the third best quality you can get with that codec and not the very best quality. When you handicap an HEVC encoder in this way, it's not surprising that you get such poor results. Dror Gill: 11:02 I think based on all these points everybody can understand why the results of this comparison were quite different than all of the other comparison that were published a month earlier at IBC (by Ateme, BBC, Harmonic). Mark Donnigan: 11:13 It's interesting. Mark Donnigan: 11:14 Another critical topic that we have to cover is performance. If you measure the CPU performance on encoding time of AV1, I believe that it's pretty universally understood that you are going to find it currently is a hundred times slower than HEVC. Is that correct? Dror Gill: 11:32 Yeah, that's right. Typically, you measure the performance of an encoder and FPS which is frames per second. For HEVC it's common to measure an FPM which is frames per minute. Mark Donnigan: 11:42 Frames per minute, (more like) frames per hour, FPH. Dror Gill: 11:45 A year and a half ago or a year ago when there were very initial implementation, it was really FPD or FPH, Frames per hour or per day and you really needed to have a lot of patience, but now after they've done some work it's only a hundred times slower than HEVC. Mark Donnigan: 12:02 Yeah, that's pretty good. They're getting there. But some people say that the open source implementation of AV1 I believe it's AOM ENC. Dror Gill: 12:11 Yeah, AOM ENC. Mark Donnigan: 12:16 ENC exactly has not been optimized for performance at all. One thing I like about speed is either your encoder produces X number of frames per second or per minute, or it doesn't. It's really simple. Here is my next question for you. Proponents of AV1 are saying, “well it's true it's slow but it hasn't been optimized, the open source implementation,” which is to imply that there's a lot of room (for improvement) and that we're just getting started, “don't worry we'll close the gap.” But if you look at the code, and by the way I may be a marketing guy but my formal education is computer science. Mark Donnigan: 13:03 You can see it already includes performance optimizations. I mean eptimizations like MMX, SSE, there's AVX instructions, there's CPU optimization, there's multithreading. It seems like they're already trying to make this thing go faster. So how are they going to close this a hundred X (time) gap? Dror Gill: 13:22 I don't think they can. I mean a hundred X, that's a lot and you know even the AV1 guys they even admit that they won't be able to close the gap. I talked to a few senior people who're involved in the Alliance for Open Media and even they told me that they expect AV1 to five to 10 times more complex than HEVC at the end of the road. In two to three years after all optimization are done, it's still going to be more complex than HEVC. Dror Gill: 13:55 Now, if you ask me why it's so complex I'll tell you my opinion. Okay, this is my personal opinion. I think it's because they invested a lot of effort in side stepping the patents (HEVC). Mark Donnigan: 14:07 Good point. I agree. Dror Gill: 14:07 They need to get that compression efficiency which is the same as HEVC but they need to use algorithms that are not patented. They have methods that use much more CPU resources than the original patent algorithms to reach the same results. You can call it kind of brute force implementation of the same thing to avoid the patent issue. That's my personal opinion, but the end result I think is clear, it's going to be five to 10 times slower than HEVC. It has the same compression efficiency so I think it's quite questionable. This whole notion of using AV1 to get better results. Mark Donnigan: 14:45 Absolutely. If you can encode let's say on a single computer with HEVC a full ABR stack, this is what people want to do. But here we're talking speeds that are so slow let's just try and do (encode) one stream. Literally what you're saying is you'll need five to 10 computers to do the same encode with AV1. I mean, that's just not viable. It doesn't make sense to me. Dror Gill: 15:14 Yeah, why would you invest so much encoding into getting the same results. If you look at another aspect of this, let's talk about hardware encode. Companies that have large data centers, companies that are encoding vast amount of video content are not looking into moving from the traditional software encoding and CPUs and GPUs, to dedicated hardware. We're hearing talks about FPGAs even ASICs … by the way this is a very interesting trend in itself that we'll probably cover in one of the next episodes. But in the context of AV1, imagine a chip that is five to 10 times larger than an HEVC chip and which is the same complexity efficiency. The question I ask again is why? Why would anybody design such a chip, and why would anybody use it when HEVC is available today? It's much easier to encode, royalty issues have been practically solved so you know? Mark Donnigan: 16:06 Yeah, it's a big mystery for sure. One thing I can say is the Alliance for Open Media has done a great service to HEVC by pushing the patent holders to finalize their licensing terms … and ultimately make them much more rational shall we say? Dror Gill: 16:23 Yeah. Mark Donnigan: 16:25 Let me say that as we're an HEVC vendor and speaking on behalf of others (in the industry), we're forever thankful to the Alliance for Open Media. Dror Gill: 16:36 Definitely, without the push from AOM and the development of AV1 we would be stuck with HEVC royalty issue until this day. Mark Donnigan: 16:44 That was not a pretty situation a few years back, wow! Dror Gill: 16:48 No, no, but as we said in the last episode we have a “happy ending” now. (reference to episode 1) Mark Donnigan: 16:52 That's right. Dror Gill: 16:52 Billions of devices support HEVC and royalty issues are pretty much solved, so that's great. I think we've covered HEVC and AV1 pretty thoroughly in two episodes but what about the other codecs? There's VP9, you could call that the predecessor of AV1, and then there's VVC, which is the successor of HEVC. It's the next codec developed by MPEG. Okay, VP9 and VVC I guess we have a topic for our next episode, right? Mark Donnigan: 17:21 It's going to be awesome. Narrator: 17:23 Thank you for listening to the Video Insider podcast a production of Beamr limited. To begin using Beamr codecs today go to beamr.com/free to receive up to 100 hours of no cost HEVC and H.264 transcoding every month.
The 15th Annual Symposium on Communication in 2015 kicks off at Baruch College campus. The second speaker is Manuel Lima, Founder of visualcomplexity.com and author of Book of Trees: Visualizing Branches of Knowledge. The topic of Mr. Lima's lecture focuses on the information visualization and the phenomenon of transformation from data to information, knowledge, and wisdom.
Professor Weierich leads the presentation of this lecture to discuss about the visual scenes and its connection with the eye tracking tasks in marketing, sociology, psychology, and other fields.