Max Seitzer
@maxseitzer.bsky.social
Research Scientist in the DINO team at Meta FAIR. Previously: PhD at Max-Planck Institute for Intelligent Systems, Tübingen. Representation learning, agents, structure.
Pinned
Max Seitzer
@maxseitzer.bsky.social
· Aug 14
Introducing DINOv3 🦕🦕🦕
A SotA-enabling vision foundation model, trained with pure self-supervised learning (SSL) at scale.
High quality dense features, combining unprecedented semantic and geometric scene understanding.
Three reasons why this matters👇
A SotA-enabling vision foundation model, trained with pure self-supervised learning (SSL) at scale.
High quality dense features, combining unprecedented semantic and geometric scene understanding.
Three reasons why this matters👇
Introducing DINOv3 🦕🦕🦕
A SotA-enabling vision foundation model, trained with pure self-supervised learning (SSL) at scale.
High quality dense features, combining unprecedented semantic and geometric scene understanding.
Three reasons why this matters👇
A SotA-enabling vision foundation model, trained with pure self-supervised learning (SSL) at scale.
High quality dense features, combining unprecedented semantic and geometric scene understanding.
Three reasons why this matters👇
August 14, 2025 at 6:52 PM
Introducing DINOv3 🦕🦕🦕
A SotA-enabling vision foundation model, trained with pure self-supervised learning (SSL) at scale.
High quality dense features, combining unprecedented semantic and geometric scene understanding.
Three reasons why this matters👇
A SotA-enabling vision foundation model, trained with pure self-supervised learning (SSL) at scale.
High quality dense features, combining unprecedented semantic and geometric scene understanding.
Three reasons why this matters👇
Reposted by Max Seitzer
✨Introducing SENSEI✨ We bring semantically meaningful exploration to model-based RL using VLMs.
With intrinsic rewards for novel yet useful behaviors, SENSEI showcases strong exploration in MiniHack, Pokémon Red & Robodesk.
Accepted at ICML 2025🎉
Joint work with @cgumbsch.bsky.social
🧵
With intrinsic rewards for novel yet useful behaviors, SENSEI showcases strong exploration in MiniHack, Pokémon Red & Robodesk.
Accepted at ICML 2025🎉
Joint work with @cgumbsch.bsky.social
🧵
July 14, 2025 at 8:02 AM
✨Introducing SENSEI✨ We bring semantically meaningful exploration to model-based RL using VLMs.
With intrinsic rewards for novel yet useful behaviors, SENSEI showcases strong exploration in MiniHack, Pokémon Red & Robodesk.
Accepted at ICML 2025🎉
Joint work with @cgumbsch.bsky.social
🧵
With intrinsic rewards for novel yet useful behaviors, SENSEI showcases strong exploration in MiniHack, Pokémon Red & Robodesk.
Accepted at ICML 2025🎉
Joint work with @cgumbsch.bsky.social
🧵
Reposted by Max Seitzer
Scaling 4D Representations
Self-supervised learning from video does scale! In our latest work, we scaled masked auto-encoding models to 22B params, boosting performance on pose estimation, tracking & more.
Paper: arxiv.org/abs/2412.15212
Code & models: github.com/google-deepmind/representations4d
Self-supervised learning from video does scale! In our latest work, we scaled masked auto-encoding models to 22B params, boosting performance on pose estimation, tracking & more.
Paper: arxiv.org/abs/2412.15212
Code & models: github.com/google-deepmind/representations4d
July 10, 2025 at 11:52 AM
Scaling 4D Representations
Self-supervised learning from video does scale! In our latest work, we scaled masked auto-encoding models to 22B params, boosting performance on pose estimation, tracking & more.
Paper: arxiv.org/abs/2412.15212
Code & models: github.com/google-deepmind/representations4d
Self-supervised learning from video does scale! In our latest work, we scaled masked auto-encoding models to 22B params, boosting performance on pose estimation, tracking & more.
Paper: arxiv.org/abs/2412.15212
Code & models: github.com/google-deepmind/representations4d
Reposted by Max Seitzer
Introducing 3DGSim🧩— an end-to-end 3D physics simulator trained only on multi-view videos. It achieves spatial & temporal consistency w/o ground truth 3D info or heavy inductive biases— enabling scalability & generalization🚀
Kudos to Mikel + @andregeist.bsky.social
www.youtube.com/watch?v=3Ar3...
Kudos to Mikel + @andregeist.bsky.social
www.youtube.com/watch?v=3Ar3...
mikel-zhobro.github.io
April 4, 2025 at 9:08 AM
Introducing 3DGSim🧩— an end-to-end 3D physics simulator trained only on multi-view videos. It achieves spatial & temporal consistency w/o ground truth 3D info or heavy inductive biases— enabling scalability & generalization🚀
Kudos to Mikel + @andregeist.bsky.social
www.youtube.com/watch?v=3Ar3...
Kudos to Mikel + @andregeist.bsky.social
www.youtube.com/watch?v=3Ar3...