Prev play stop Next mute max volume 00:00 00:00 repeat Update Required To play the media you will need to either update your browser to a recent version or update your Flash plugin. OpenAI CLIP: ConnectingText and Images (Paper Explained) Yannic Kilcher Kho Tổng Hợp 173,224 5 năm trước Xem video Facebook Tweet XEM MÔ TẢ #ai #openai #technology Paper Title: Learning Transferable Visual Models From Natural Language Supervision CLIP trains on 400 million images scraped from the web, along with text descriptions to learn a model that can connect the two modalities. The core idea is a contrastive objective combined with a large batch size. The resulting model can be turned into arbitrary zero-shot classifiers for new image & text tasks. OUTLINE: 0:00 - Introduction 3:15 - Overview 4:40 - Connecting Images & Text 9:00 - Building Zero-Shot Classifiers 14:40 - CLIP Contrastive Training Objective 22:25 - Encoder Choices 25:00 - Zero-Shot CLIP vs Linear ResNet-50 31:50 - Zero-Shot vs Few-Shot 35:35 - Scaling Properties 36:35 - Comparison on different tasks 37:40 - Robustness to Data Shift 44:20 - Broader Impact Section 47:00 - Conclusion & Comments Paper: https://cdn.openai.com/papers/Learning_Transferable_Visual_Models_From_Natural_Language_Supervision.pdf Blog: https://openai.com/blog/clip/ Code: https://github.com/openai/CLIP Abstract: State-of-the-art computer vision systems are trained to predict a fixed set of predetermined object categories. This restricted form of supervision limits their generality and usability since additional labeled data is needed to specify any other visual concept. Learning directly from raw text about images is a promising alternative which leverages a much broader source of supervision. We demonstrate that the simple pre-training task of predicting which caption goes with which image is an efficient and scalable way to learn SOTA image representations from scratch on a dataset of 400 million (image, text) pairs collected from the internet. After pre-training, natural language is used to reference learned visual concepts (or describe new ones) enabling zero-shot transfer of the model to downstream tasks. We study the performance of this approach by benchmarking on over 30 different existing computer vision datasets, spanning tasks such as OCR, action recognition in videos, geo-localization, and many types of fine-grained object classification. The model transfers non-trivially to most tasks and is often competitive with a fully supervised baseline without the need for any dataset specific training. For instance, we match the accuracy of the original ResNet-50 on ImageNet zero-shot without needing to use any of the 1.28 million training examples it was trained on. Authors: Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, Ilya Sutskever Links: TabNine Code Completion (Referral): http://bit.ly/tabnine-yannick YouTube: https://www.youtube.com/c/yannickilcher Twitter: https://twitter.com/ykilcher Discord: https://discord.gg/4H8xxDF BitChute: https://www.bitchute.com/channel/yannic-kilcher Minds: https://www.minds.com/ykilcher Parler: https://parler.com/profile/YannicKilcher LinkedIn: https://www.linkedin.com/in/yannic-kilcher-488534136/ BiliBili: https://space.bilibili.com/1824646584 If you want to support me, the best thing to do is to share out the content :) If you want to support me financially (completely optional and voluntary, but a lot of people have asked for this): SubscribeStar: https://www.subscribestar.com/yannickilcher Patreon: https://www.patreon.com/yannickilcher Bitcoin (BTC): bc1q49lsw3q325tr58ygf8sudx2dqfguclvngvy2cq Ethereum (ETH): 0x7ad3513E3B8f66799f507Aa7874b1B0eBC7F85e2 Litecoin (LTC): LQW2TRyKYetVC8WjFkhpPhtpbDM4Vw7r9m Monero (XMR): 4ACL8AGrEo5hAir8A9CeVrW8pEauWvnp1WnSDZxW7tziCDLhZAGsgzhRQABDnFy8yuM9fWJDviJPHKRjV4FWt19CJZN9D4n Mp3 liên quan 12:23 Cách Làm Xíu Mại Bán Bánh Mì Thơm Ngon-Hướng Dẫn làm món Xíu Mại-Món Ngon Mẹ Nấu Món Ngon Mẹ Nấu Mom's Best Food 1,345,018 view 6 năm trước 0:16 eye contact challenge with BTS v...❤️#bighitofficial#bts#army#bts_purple_world#kimtaehyung#v#tiger 💜BTS Purple World💜 732,129 view 1 năm trước 1:55 Meet Tom McDougall, Founder and CEO, 4P Foods | J.P. Morgan jpmorgan 500 view 2 năm trước 24:48 Quán cơm xe Tải bắc Nam có món ngon bí ẩn khiến nhiều người mê Tuấn container 138,361 view 6 tháng trước 27:12 ALL ACCESS: Spence vs. Crawford | Epilogue | Full Episode | SHOWTIME PPV ALL THE SMOKE FIGHT 3,364,903 view 2 năm trước 0:56 The Evolution of French Women’s Fashion 🇫🇷 Majestic Studios 7,582,251 view 10 tháng trước 2:01:21 Liên Khúc Trữ Tình Bolero Đang Được Yêu Thích - Người Không Cô Đơn | Phương Anh Phương Anh Bolero 218,359 view 1 năm trước 1:11 How Does Adrenaline Prepare Your Body for Action? Simple Science 1,022 view 1 năm trước 6:46 Cách làm CỦ KIỆU NGÂM CHUA NGỌT từ ĐƯỜNG Trắng Cực Giòn và để được lâu Nắng Vào Bếp 105,252 view 1 năm trước 52:18 TÌNH KHÚC TIỀN CHIẾN BẤT HỦ -tiếng hát MỸ HẠNH TIẾNG HÁT MỸ HẠNH 318,030 view 9 tháng trước 0:58 Fantastic Four Movie Theory Explained #shorts ScreenRant 7,661 view 1 năm trước 6:53 Cách làm Đu Đủ Ngâm Chua Ngọt giòn rụm ngon hơn nhiều so với dưa chua củ cải-ALO TRÀ VINH ALO TRÀ VINH 91,757 view 5 năm trước 2:02:18 Nhạc BOLERO TRỮ TÌNH HAY, MỚI NHẤT 2026 - Nhạc Vàng Sến Xưa Vừa Ra Lò DỄ NGHE & XOA DỊU CON TIM Bolero Phố Cũ 481,634 view 6 tháng trước 9:57 Did Voicing Princess Peach Make Anya Taylor Joy VOMIT?! She Clears Up Viral Super Mario Story Entertainment Tonight 20,318 view 3 tháng trước 12:37 Chả Lụa Chay Từ Đậu Hủ, Dai Mịn Không Bột Ngon Hơn Thực Dưỡng Mỗi Ngày 50,272 view 1 năm trước 9:53 Heo Quay Da Giòn - Crispy Pork Belly - Chia sẻ kinh nghiệm làm heo quay DA GIÒN RỤM Diễm Nauy 626,723 view 5 năm trước 29:22 ALL ACCESS: Davis vs. Barrios | Epilogue | Full Episode (TV14) | SHOWTIME PPV ALL THE SMOKE FIGHT 4,707,385 view 5 năm trước 11:51 How to ACTUALLY Get Better At Apex HisWattson 156,524 view 8 tháng trước 5:27 Lên Kèo Đi Ăn Món MÌ TRỘN CHẢ CUA Ăn Là Ghiền Thơm Ngon Chỉ Có Ở Phú Nhuận | Địa điểm ăn uống Địa điểm ăn uống 54,063 view 3 năm trước 0:48 Hilarious Anya Taylor-Joy & Nicholas Hoult Bloopers 😂 IGV Presents 7,731 view 2 năm trước Thể loại nhạc