Generative AI’s Post

𝐌𝐚𝐫𝐤 𝐙𝐮𝐜𝐤𝐞𝐫𝐛𝐞𝐫𝐠 𝐉𝐮𝐬𝐭 𝐀𝐧𝐧𝐨𝐮𝐧𝐜𝐞𝐝 𝐌𝐞𝐭𝐚 𝐒𝐀𝐌 𝟐: 𝐑𝐞𝐚𝐥-𝐓𝐢𝐦𝐞 𝐎𝐛𝐣𝐞𝐜𝐭 𝐒𝐞𝐠𝐦𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥 ...which is more accurate and six times faster than the original SAM 𝐖𝐡𝐚𝐭’𝐬 𝐍𝐞𝐰? ✅ SAM 2 works with video where the original SAM only worked with still images ✅ It outperforms SAM on its 23 dataset zero-shot benchmark suite, while outperforms other approaches across 17 zero-shot video datasets ✅ The new SA-V dataset is a game-changer, 4.5 times larger than previous datasets with 53 times more annotations. (51,000 videos and over 600,000 spatio-temporal masks) ✅ Released under the Apache 2.0 license, meaning anyone can use it to create innovative applications. Meta said the AI model can help ease the process of video editing or AI-based video generation, as well as to power new experiences in the company's mixed-reality ecosystem. Official website: https://lnkd.in/dtNadw87 Project demo page: https://lnkd.in/dcBr6Bnu Model download: https://lnkd.in/dkWDB7kN What do you think about SAM 2? How do you see it impacting your field? #ai #tech #innovation

𝐌𝐚𝐫𝐤 𝐙𝐮𝐜𝐤𝐞𝐫𝐛𝐞𝐫𝐠 𝐉𝐮𝐬𝐭 𝐀𝐧𝐧𝐨𝐮𝐧𝐜𝐞𝐝 𝐌𝐞𝐭𝐚 𝐒𝐀𝐌 𝟐: 𝐑𝐞𝐚𝐥-𝐓𝐢𝐦𝐞 𝐎𝐛𝐣𝐞𝐜𝐭 𝐒𝐞𝐠𝐦𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥 ...which is more accurate and six times faster than the original SAM 𝐖𝐡𝐚𝐭’𝐬 𝐍𝐞𝐰? ✅ SAM 2 works with video where the original SAM only worked with still images ✅ It outperforms SAM on its 23 dataset zero-shot benchmark suite, while outperforms other approaches across 17 zero-shot video datasets ✅ The new SA-V dataset is a game-changer, 4.5 times larger than previous datasets with 53 times more annotations. (51,000 videos and over 600,000 spatio-temporal masks) ✅ Released under the Apache 2.0 license, meaning anyone can use it to create innovative applications. Meta said the AI model can help ease the process of video editing or AI-based video generation, as well as to power new experiences in the company's mixed-reality ecosystem. Official website: https://lnkd.in/dtNadw87 Project demo page: https://lnkd.in/dcBr6Bnu Model download: https://lnkd.in/dkWDB7kN What do you think about SAM 2? How do you see it impacting your field? #ai #tech #innovation

𝐌𝐚𝐫𝐤 𝐙𝐮𝐜𝐤𝐞𝐫𝐛𝐞𝐫𝐠 𝐉𝐮𝐬𝐭 𝐀𝐧𝐧𝐨𝐮𝐧𝐜𝐞𝐝 𝐌𝐞𝐭𝐚 𝐒𝐀𝐌 𝟐: 𝐑𝐞𝐚𝐥-𝐓𝐢𝐦𝐞 𝐎𝐛𝐣𝐞𝐜𝐭 𝐒𝐞𝐠𝐦𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥 ...which is more accurate and six times faster than the original SAM 𝐖𝐡𝐚𝐭’𝐬 𝐍𝐞𝐰? ✅ SAM 2 works with video where the original SAM only worked with still images ✅ It outperforms SAM on its 23 dataset zero-shot benchmark suite, while outperforms other approaches across 17 zero-shot video datasets ✅ The new SA-V dataset is a game-changer, 4.5 times larger than previous datasets with 53 times more annotations. (51,000 videos and over 600,000 spatio-temporal masks) ✅ Released under the Apache 2.0 license, meaning anyone can use it to create innovative applications. Meta said the AI model can help ease the process of video editing or AI-based video generation, as well as to power new experiences in the company's mixed-reality ecosystem. Official website: https://lnkd.in/dtNadw87 Project demo page: https://lnkd.in/dcBr6Bnu Model download: https://lnkd.in/dkWDB7kN What do you think about SAM 2? How do you see it impacting your field? #ai #tech #innovation

💥💥FeatUp: Any Model at Any Resolution💥💥 👉FeatUp is a task- and model-agnostic framework to restore lost spatial information in deep features. It outperforms other feature upsampling in class activation map generation, transfer learning for segmentation and depth prediction, and end-to-end training for semantic segm. Source Code released💙 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬: ✅Improving spatial res. of any model’s feats. ✅Fast #CUDA Joint Bilateral Upsampling ✅Orders of magnitude faster than #PyTorch ✅Up to 32x SR of any model's features ✅drop-in feats replacements to improve performance ✅Authors: Berkeley, MIT, #Microsoft, #Adobe, & #Google #artificialintelligence #machinelearning #ml #AI #deeplearning #computervision #AIwithPapers #metaverse 👉Discussion https://lnkd.in/dMgakzWm 👉Paper https://lnkd.in/gweaN4s6 👉Project https://lnkd.in/gWcGXdxt 👉Code https://lnkd.in/gweq5NY4

💥💥FeatUp: Any Model at Any Resolution💥💥 👉FeatUp is a task- and model-agnostic framework to restore lost spatial information in deep features. It outperforms other feature upsampling in class activation map generation, transfer learning for segmentation and depth prediction, and end-to-end training for semantic segm. Source Code released💙 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬: ✅Improving spatial res. of any model’s feats. ✅Fast #CUDA Joint Bilateral Upsampling ✅Orders of magnitude faster than #PyTorch ✅Up to 32x SR of any model's features ✅drop-in feats replacements to improve performance ✅Authors: Berkeley, MIT, #Microsoft, #Adobe, & #Google #artificialintelligence #machinelearning #ml #AI #deeplearning #computervision #AIwithPapers #metaverse 👉Discussion https://lnkd.in/dMgakzWm 👉Paper https://lnkd.in/gweaN4s6 👉Project https://lnkd.in/gWcGXdxt 👉Code https://lnkd.in/gweq5NY4

A hard part of building AI applications is choosing which model to use. What if we don’t have to? What if we can predict the best model for any prompt? Predictive human preference aims to predict which model users might prefer for a specific query. https://lnkd.in/gxy3KYGG One use case is model routing. If we know in advance that for a prompt, users will prefer Claude Instant’s response over GPT-4, and Claude Instant is cheaper/faster than GPT-4, we can route this prompt to Claude Instant. Model routing has the potential to increase response quality while reducing costs and latency. One pattern is that for simple prompts, weak models can do (nearly) as well as strong models. For more challenging prompts, however, users are much more likely to prefer stronger models. Here’s a visualization of predicted human preference for an easy prompt (“hello, how are you?”) and a challenging prompt (“Explain why Planc length …”). Preference predictors make it possible to create leaderboards unique to any prompt and domain. As always, feedback is much appreciated! #aiengineering #llm #aiapplications

A hard part of building AI applications is choosing which model to use. What if we don’t have to? What if we can predict the best model for any prompt? Predictive human preference aims to predict which model users might prefer for a specific query. https://lnkd.in/gxy3KYGG One use case is model routing. If we know in advance that for a prompt, users will prefer Claude Instant’s response over GPT-4, and Claude Instant is cheaper/faster than GPT-4, we can route this prompt to Claude Instant. Model routing has the potential to increase response quality while reducing costs and latency. One pattern is that for simple prompts, weak models can do (nearly) as well as strong models. For more challenging prompts, however, users are much more likely to prefer stronger models. Here’s a visualization of predicted human preference for an easy prompt (“hello, how are you?”) and a challenging prompt (“Explain why Planc length …”). Preference predictors make it possible to create leaderboards unique to any prompt and domain. As always, feedback is much appreciated! #aiengineering #llm #aiapplications

A hard part of building AI applications is choosing which model to use. What if we don’t have to? What if we can predict the best model for any prompt? Predictive human preference aims to predict which model users might prefer for a specific query. https://lnkd.in/gxy3KYGG One use case is model routing. If we know in advance that for a prompt, users will prefer Claude Instant’s response over GPT-4, and Claude Instant is cheaper/faster than GPT-4, we can route this prompt to Claude Instant. Model routing has the potential to increase response quality while reducing costs and latency. One pattern is that for simple prompts, weak models can do (nearly) as well as strong models. For more challenging prompts, however, users are much more likely to prefer stronger models. Here’s a visualization of predicted human preference for an easy prompt (“hello, how are you?”) and a challenging prompt (“Explain why Planc length …”). Preference predictors make it possible to create leaderboards unique to any prompt and domain. As always, feedback is much appreciated! #aiengineering #llm #aiapplications