正在加载视频...

视频加载失败

返回首页

I'll always root for a team that open-sources its best work, and Robbyant just did it properly. Robbyant, Ant Group's embodied-AI company, released LingBot-Vision, a vision foundation model for robots, and the part I love is the data. They trained it on 161M images, filtered down from 2B raw...

kimmonismus's profile picture

Chubby♨️

124,002 subscribers

37,953 次观看 • 10 小时前 •via X (Twitter)

科学技术健康养生教育
LIVE

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

You can't 3D reconstruct glass from images... ...WRONG! Thanks for video diffusion, now just about anything is possible! Introducing...Diffusion Knows Transparency (DKT) Transparent and reflective objects usually break robot vision and photogrammetry pipelines because they don't follow the "solid object" rules standard cameras expect. DKT is a new AI model that repurposes the "internal physics engine" found in video generation models to solve this problem. Researchers took a massive video diffusion model (WAN) and fine-tuned it using a custom-built synthetic dataset to turn it into a high-precision depth sensor. To train the AI, they built the first massive synthetic video library of transparent objects, 1.32 million frames of perfectly labeled glass and metal objects in motion. Without ever seeing a "real" labeled video of glass during training, the model (DKT) outperformed all previous specialized systems on real-world benchmarks (ClearPose, DREDS). They created a "lightweight" 1.3B parameter version that runs fast enough (0.17s per frame) to be used on actual robot hardware. Two reasons I find this project important: 1. It further proves that synthetic data will be essential for training the next generation vision models. 2. In real-world robotic tests, using DKT's depth maps nearly doubled the success rate of robot arms trying to pick up objects on tricky reflective or translucent surfaces. At home robots will need to interact with these types of objects on a daily basis. Check out the project page here: Code is LIVE! #Computervision #Robotics #AI
0:14

You can't 3D reconstruct glass from images... ...WRONG! Thanks for video diffusion, now just about anything is possible! Introducing...Diffusion Knows Transparency (DKT) Transparent and reflective objects usually break robot vision and photogrammetry pipelines because they don't follow the "solid object" rules standard cameras expect. DKT is a new AI model that repurposes the "internal physics engine" found in video generation models to solve this problem. Researchers took a massive video diffusion model (WAN) and fine-tuned it using a custom-built synthetic dataset to turn it into a high-precision depth sensor. To train the AI, they built the first massive synthetic video library of transparent objects, 1.32 million frames of perfectly labeled glass and metal objects in motion. Without ever seeing a "real" labeled video of glass during training, the model (DKT) outperformed all previous specialized systems on real-world benchmarks (ClearPose, DREDS). They created a "lightweight" 1.3B parameter version that runs fast enough (0.17s per frame) to be used on actual robot hardware. Two reasons I find this project important: 1. It further proves that synthetic data will be essential for training the next generation vision models. 2. In real-world robotic tests, using DKT's depth maps nearly doubled the success rate of robot arms trying to pick up objects on tricky reflective or translucent surfaces. At home robots will need to interact with these types of objects on a daily basis. Check out the project page here: Code is LIVE! #Computervision #Robotics #AI

Jonathan Stephens

17,712 次观看 • 6 个月前

Depth Any Video with Scalable Synthetic Data AI physicists and chemists continue to make strides in depth estimation from video. Check out this new paper featuring some impressive examples. See the thread for more details (unfortunately no code yet). Abstract: Video depth estimation has long been hindered by the scarcity of consistent and scalable ground truth data, leading to inconsistent and unreliable results. In this paper, we introduce Depth Any Video, a model that tackles the challenge through two key innovations. First, we develop a scalable synthetic data pipeline, capturing real-time video depth data from diverse game environments, yielding 40,000 video clips of 5-second duration, each with precise depth annotations. Second, we leverage the powerful priors of generative video diffusion models to handle real-world videos effectively, integrating advanced techniques such as rotary position encoding and flow matching to further enhance flexibility and efficiency. Unlike previous models, which are limited to fixed-length video sequences, our approach introduces a novel mixed-duration training strategy that handles videos of varying lengths and performs robustly across different frame rates 0 - even on single frames. At inference, we propose a depth interpolation method that enables our model to infer high-resolution video depth across sequences of up to 150 frames. Our model outperforms all previous generative depth models in terms of spatial accuracy and temporal consistency.
0:13

Depth Any Video with Scalable Synthetic Data AI physicists and chemists continue to make strides in depth estimation from video. Check out this new paper featuring some impressive examples. See the thread for more details (unfortunately no code yet). Abstract: Video depth estimation has long been hindered by the scarcity of consistent and scalable ground truth data, leading to inconsistent and unreliable results. In this paper, we introduce Depth Any Video, a model that tackles the challenge through two key innovations. First, we develop a scalable synthetic data pipeline, capturing real-time video depth data from diverse game environments, yielding 40,000 video clips of 5-second duration, each with precise depth annotations. Second, we leverage the powerful priors of generative video diffusion models to handle real-world videos effectively, integrating advanced techniques such as rotary position encoding and flow matching to further enhance flexibility and efficiency. Unlike previous models, which are limited to fixed-length video sequences, our approach introduces a novel mixed-duration training strategy that handles videos of varying lengths and performs robustly across different frame rates 0 - even on single frames. At inference, we propose a depth interpolation method that enables our model to infer high-resolution video depth across sequences of up to 150 frames. Our model outperforms all previous generative depth models in terms of spatial accuracy and temporal consistency.

MrNeRF

27,428 次观看 • 1 年前

Some random thoughts reading through the new RobbyAnt VLA paper: - 20,000 hours of data across 9 robots!! - damn, Chinese companies are going to trivially outscale the American ones on real robot data - It's really cool they train on depth; it means they can handle transparent objects really well for example - You can never tell how good these models are without trying them, since everyone trains on different robots, but from the results they show it does seem to clean up - cross embodiment scaling laws are really cool - if you have lots of robots do you need human video??
0:13

Some random thoughts reading through the new RobbyAnt VLA paper: - 20,000 hours of data across 9 robots!! - damn, Chinese companies are going to trivially outscale the American ones on real robot data - It's really cool they train on depth; it means they can handle transparent objects really well for example - You can never tell how good these models are without trying them, since everyone trains on different robots, but from the results they show it does seem to clean up - cross embodiment scaling laws are really cool - if you have lots of robots do you need human video??

Chris Paxton

19,921 次观看 • 5 个月前

When the lens of a camera tilts so that it is no longer parallel to the camera's image sensor, it shifts the plane of focus and alters the depth of field. This is a tilt shift view of farming operations. [📹 albaxya0]
0:15

When the lens of a camera tilts so that it is no longer parallel to the camera's image sensor, it shifts the plane of focus and alters the depth of field. This is a tilt shift view of farming operations. [📹 albaxya0]

Massimo

1,626,807 次观看 • 2 年前

🧵 On food in Gaza: One of the best and most reliable ways to find out what’s actually happening on the ground in Gaza is through open source data - such as Snapchat, where Palestinians share stories from their everyday life. It shows which kind of food they receive and which countries it comes from. The following videos and pictures are from the last couple of days in Gaza City, Khan Younis and Al Mawasi, to name a few places.
0:27

🧵 On food in Gaza: One of the best and most reliable ways to find out what’s actually happening on the ground in Gaza is through open source data - such as Snapchat, where Palestinians share stories from their everyday life. It shows which kind of food they receive and which countries it comes from. The following videos and pictures are from the last couple of days in Gaza City, Khan Younis and Al Mawasi, to name a few places.

Jotam Confino

22,729 次观看 • 10 个月前

When the lens of a camera tilts so that it is no longer parallel to the camera's image sensor, it shifts the plane of focus and alters the depth of field. This is a tilt shift view of a construction site. [📹 albaxya0]
0:17

When the lens of a camera tilts so that it is no longer parallel to the camera's image sensor, it shifts the plane of focus and alters the depth of field. This is a tilt shift view of a construction site. [📹 albaxya0]

Massimo

6,503,030 次观看 • 2 年前

A trio of mathematicians built the first physical model of a “monostable” tetrahedron, a shape that will always flip-flop onto the same face no matter what side you place it on. In order for it to work properly, it had to be engineered to a level of precision within one-tenth of a gram and one-tenth of a millimeter. (From the archive)
0:12

A trio of mathematicians built the first physical model of a “monostable” tetrahedron, a shape that will always flip-flop onto the same face no matter what side you place it on. In order for it to work properly, it had to be engineered to a level of precision within one-tenth of a gram and one-tenth of a millimeter. (From the archive)

Quanta Magazine

64,668 次观看 • 3 个月前

there is so much real data just sitting in the open right now it's almost funny. four years of starlight on every star, a NASA archive that's been free for over a decade, detectors still recording the sky tonight, and barely anyone has a net pointed at any of it. so i pointed one. this is me pulling the planet data, the data loading is the boring part. the net i built to read it, the wall it hit, and what that taught me about where AI goes next, that's the full story, and it drops tonight. the data's public, the tools are free, the box fits on a desk. what's stopping you. you can just do things anon.
0:11

there is so much real data just sitting in the open right now it's almost funny. four years of starlight on every star, a NASA archive that's been free for over a decade, detectors still recording the sky tonight, and barely anyone has a net pointed at any of it. so i pointed one. this is me pulling the planet data, the data loading is the boring part. the net i built to read it, the wall it hit, and what that taught me about where AI goes next, that's the full story, and it drops tonight. the data's public, the tools are free, the box fits on a desk. what's stopping you. you can just do things anon.

Sudo su

60,445 次观看 • 1 个月前

Introducing ml-intern, the agent that just automated the post-training team Hugging Face It's an open-source implementation of the real research loop that our ML researchers do every day. You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem. It can pull off crazy things: We made it train the best model for scientific reasoning. It went through citations from the official benchmark paper. Found OpenScience and NemoTron-CrossThink, added 7 difficulty-filtered dataset variants from ARC/SciQ/MMLU, and ran 12 SFT runs on Qwen3-1.7B. This pushed the score 10% → 32% on GPQA in under 10h. Claude Code's best: 22.99%. In healthcare settings it inspected available datasets, concluded they were too low quality, and wrote a script to generate 1100 synthetic data points from scratch for emergencies, hedging, multilingual etc. Then upsampled 50x for training. Beat Codex on HealthBench by 60%. For competitive mathematics, it wrote a full GRPO script, launched training with A100 GPUs on watched rewards claim and then collapse, and ran ablations until it succeeded. All fully backed by papers, autonomously. How it works? ml-intern makes full use of the HF ecosystem: - finds papers on arxiv and reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on - browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data - launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like. Releasing it today as a CLI and a web app you can use from your phone/desktop. CLI: Web + mobile: And the best part? We also provisioned 1k$ GPU resources and Anthropic credits for the quickest among you to use.
0:16

Introducing ml-intern, the agent that just automated the post-training team Hugging Face It's an open-source implementation of the real research loop that our ML researchers do every day. You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem. It can pull off crazy things: We made it train the best model for scientific reasoning. It went through citations from the official benchmark paper. Found OpenScience and NemoTron-CrossThink, added 7 difficulty-filtered dataset variants from ARC/SciQ/MMLU, and ran 12 SFT runs on Qwen3-1.7B. This pushed the score 10% → 32% on GPQA in under 10h. Claude Code's best: 22.99%. In healthcare settings it inspected available datasets, concluded they were too low quality, and wrote a script to generate 1100 synthetic data points from scratch for emergencies, hedging, multilingual etc. Then upsampled 50x for training. Beat Codex on HealthBench by 60%. For competitive mathematics, it wrote a full GRPO script, launched training with A100 GPUs on watched rewards claim and then collapse, and ran ablations until it succeeded. All fully backed by papers, autonomously. How it works? ml-intern makes full use of the HF ecosystem: - finds papers on arxiv and reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on - browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data - launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like. Releasing it today as a CLI and a web app you can use from your phone/desktop. CLI: Web + mobile: And the best part? We also provisioned 1k$ GPU resources and Anthropic credits for the quickest among you to use.

Aksel

1,263,557 次观看 • 2 个月前

Across the open web, vast unstructured data is generated 24/7. From text and images to videos and streams. In a world of AI agents that create content or analyze markets, nearly all depend on one essential ingredient: fresh, live, unfiltered data from the open web. Teneo Protocol delivers exactly that. Provides the infrastructure others build on.
0:14

Across the open web, vast unstructured data is generated 24/7. From text and images to videos and streams. In a world of AI agents that create content or analyze markets, nearly all depend on one essential ingredient: fresh, live, unfiltered data from the open web. Teneo Protocol delivers exactly that. Provides the infrastructure others build on.

Teneo Protocol

78,765 次观看 • 9 个月前

🎉 The web version of the Midjourney Styles Collection is live 1200+ people who purchased it on Gumroad (link in the thread) just received an email from me with the link and a password. This is an MVP, and I'm open to feedback and feature requests. Have a great weekend!
0:16

🎉 The web version of the Midjourney Styles Collection is live 1200+ people who purchased it on Gumroad (link in the thread) just received an email from me with the link and a password. This is an MVP, and I'm open to feedback and feature requests. Have a great weekend!

Tatiana Tsiguleva

53,889 次观看 • 1 年前

So remember that time machine I told you about? I managed to sneek out another video from my visit to the future.. this time I captured video of a 3d artist creating assets for a short commercial. They sculpted the bird in the video with nothing but two controllers which they used to manipulate generative matter in 3d. The AI model changed in realtime and adopted their style of work, both based on the work they did on the scene, and on reference images they gave the system. Seeing the 3d model transform in realtime as they were working on it was incredible. When it started moving and reacted to their instructions, I knew it was time to return to the present :D #art #ai
0:16

So remember that time machine I told you about? I managed to sneek out another video from my visit to the future.. this time I captured video of a 3d artist creating assets for a short commercial. They sculpted the bird in the video with nothing but two controllers which they used to manipulate generative matter in 3d. The AI model changed in realtime and adopted their style of work, both based on the work they did on the scene, and on reference images they gave the system. Seeing the 3d model transform in realtime as they were working on it was incredible. When it started moving and reacted to their instructions, I knew it was time to return to the present :D #art #ai

Martin Nebelong

22,044 次观看 • 2 年前

An update for everyone who is not in our Discord... Obviously we are still early in development, and our team has been working hard to release updates for everyone to see. We have a peak at the terrain, which is this picture is based of the North Dakota drift prairie, Great care, time and effort was put into the terrain to create smooth rolling hills, the road network in this region will be challenging, just as it is in real life. From Will The current state of the TIV2 model, which will include a lot more in depth details, From Jlkillen03 The stages of damage to a Blender made house, From Kunahic And how texturing will be used for assets like the interiors! From BSP
0:17

An update for everyone who is not in our Discord... Obviously we are still early in development, and our team has been working hard to release updates for everyone to see. We have a peak at the terrain, which is this picture is based of the North Dakota drift prairie, Great care, time and effort was put into the terrain to create smooth rolling hills, the road network in this region will be challenging, just as it is in real life. From Will The current state of the TIV2 model, which will include a lot more in depth details, From Jlkillen03 The stages of damage to a Blender made house, From Kunahic And how texturing will be used for assets like the interiors! From BSP

Severity

18,960 次观看 • 3 个月前

Issue 4 of Interfaces just released! This issue is about building an animated sign-in dialog. It covers everything that goes into building it, from the animations to the small but important details. It's one of the most in-depth issues yet, full of code snippets and demos.
0:15

Issue 4 of Interfaces just released! This issue is about building an animated sign-in dialog. It covers everything that goes into building it, from the animations to the small but important details. It's one of the most in-depth issues yet, full of code snippets and demos.

Jakub Krehel

36,615 次观看 • 1 个月前

Yanking and pulling through a bar at rest from the ground builds more muscle than any other timid and controlled rep IMO Many love depth drops, because of that extremely short but extremely high exposure to forces A power row is the depth jump for the upper body/back
0:13

Yanking and pulling through a bar at rest from the ground builds more muscle than any other timid and controlled rep IMO Many love depth drops, because of that extremely short but extremely high exposure to forces A power row is the depth jump for the upper body/back

Official Strength Debates

45,694 次观看 • 9 个月前

No physics libraries? No raycasts? That's right! The track is a spline that lets us easily figure out the normals and tangents at any given point on the curve. This makes collision and twisting tracks simple and super performant. Reply if you want an in-depth break down 😃
0:12

No physics libraries? No raycasts? That's right! The track is a spline that lets us easily figure out the normals and tangents at any given point on the curve. This makes collision and twisting tracks simple and super performant. Reply if you want an in-depth break down 😃

Daniel Beauchamp

39,683 次观看 • 1 年前