𝟯𝗗-𝗟𝗔𝗧𝗧𝗘: 𝗟𝗮𝘁𝗲𝗻𝘁 𝗦𝗽𝗮𝗰𝗲 𝟯𝗗 𝗘𝗱𝗶𝘁𝗶𝗻𝗴 𝗳𝗿𝗼𝗺 𝗧𝗲𝘅𝘁𝘂𝗮𝗹 𝗜𝗻𝘀𝘁𝗿𝘂𝗰𝘁𝗶𝗼𝗻𝘀 Maria Parelli, Michael Oechsle, Michael Niemeyer ... Andreas Geiger arxiv.org/abs/2509.00269 Trending on www.scholar-inbox.com

🚀 Introducing our new paper, MDPO: Overcoming the Training-Inference Divide of Masked Diffusion Language Models. 📄 Paper: www.scholar-inbox.com/papers/He202... arxiv.org/pdf/2508.13148 💻 Code: github.com/autonomousvi... 🌐 Project Page: cli212.github.io/MDPO/

Today, we moved into our new building on the CyberValley campus. Everyone is super excited. PhD students went right back to work. But wait, is there something missing? ;)

Today we had our AVG Deep Cave Expedition Day! Exploring the challenges of the (unlit, narrow, crawling-only) Hofener Höhle near Grabenstetten ..

SpatialTrackerV2: 3D Point Tracking Made Easy Yuxi Xiao, @jianyuanwang.bsky.social, Nan Xue, @nikkar.bsky.social, Yuri Makarov, Bingyi Kang, Xing Zhu, Hujun Bao, Yujun Shen, Xiaowei Zhou tl;dr: DAv2+VGGT->depths & poses->iterative cross-attention-based optimizer arxiv.org/abs/2507.12462

In case you find it as relaxing as we do: Here is a 2h+ video of our autonomous RL driving agent CaRL in action! @danieldauner.bsky.social @bernhard-jaeger.bsky.social @kashyap7x.bsky.social youtube.com/watch?v=_god...

At #ICML, you can just use scholar inbox to help you find your way through the poster sessions. It just sorts the papers according to your preferences and it really works. www.scholar-inbox.com/conference/i... ICML 2025 Planner

I am in Vancouver at ICML, and tomorrow I will present our newest paper "Partially Observable Reinforcement Learning with Memory Traces". We argue that eligibility traces are more effective than sliding windows as a memory mechanism for RL in POMDPs. 🧵

We have released the code for our work, CaRL: Learning Scalable Planning Policies with Simple Rewards. The repository contains the first public code base for training RL agents with the CARLA leaderboard 2.0 and nuPlan. github.com/autonomousvi...

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

𝗚𝗲𝗼𝗺𝗲𝘁𝗿𝘆-𝗮𝘄𝗮𝗿𝗲 𝟰𝗗 𝗩𝗶𝗱𝗲𝗼 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻 𝗳𝗼𝗿 𝗥𝗼𝗯𝗼𝘁 𝗠𝗮𝗻𝗶𝗽𝘂𝗹𝗮𝘁𝗶𝗼𝗻 Zeyi Liu, Shuang Li, Eric Cousineau ... Shuran Song arxiv.org/abs/2507.01099 Trending on www.scholar-inbox.com

MoGe-2: Accurate Monocular Geometry with Metric Scale and Sharp Details Ruicheng Wang, Sicheng Xu, Yue Dong, Yu Deng, Jianfeng Xiang, Zelong Lv, Guangzhong Sun, Xin Tong, Jiaolong Yang arxiv.org/abs/2507.02546

I am very proud of my group! These are the nationalities of my current and past team members. Diversity is key. 🇩🇪 🇬🇷 🇮🇹 🇮🇳 🇷🇺 🇺🇦 🇨🇳 🇷🇸 🇯🇵 🇧🇪 🇺🇸 🇰🇷 🇹🇷

That’s a wrap on #CVPR2025 in Nashville! From online convos to in-person vibes, one thing’s clear: this community is STRONG 💪 Thanks for following along! Until next time. @deblinaml.bsky.social, @jbhaurum.bsky.social, @csprofkgd.bsky.social signing off.

LLM product placement and search optimization is here and it's as dystopian as you expected.

Hey #CVPR2025! Curious about this work? I'll be presenting it this morning! Poster 31, from 10:30 to 12:30 🤠 @cvprconference.bsky.social

Check out the ScanNet++ workshop @CVPR on June 12 in 211 from 8:50am! Exciting keynotes on state-of-the-art NVS & 3D understanding from Andrea Vedaldi, Cordelia Schmid, Gordon Wetzstein, Katja Schwarz, Qianqian Wang, and leading methods on the benchmark! kaldir.vc.in.tum.de/scannetpp/cv...

Join us for the 4D Vision Workshop #CVPR on June 11 starting at 9:20am! We'll have an incredible lineup of speakers discussing the frontier of 3D computer vision techniques for dynamic world modeling across spatial AI, robotics, astrophysics, and more. 4dvisionworkshop.github.io

This Wednesday (1-6PM, Room 106A) at CVPR @cvprconference.bsky.social we have a great lineup of keynote speakers, posters, and spotlights on neural fields and beyond: neural-bcc.github.io Have a question you want answered by a panel of experts in the field? Send it to us via: tinyurl.com/bdddf36f

Excited to present our #CVPR2025 paper DepthSplat next week! DepthSplat is a feed-forward model that achieves high-quality Gaussian reconstruction and view synthesis in just 0.6 seconds. Looking forward to great conversations at the conference!

🚗 Pseudo-simulation combines the efficiency of open-loop and robustness of closed-loop evaluation. It uses real data + 3D Gaussian Splatting synthetic views to assess error recovery, achieving strong correlation with closed-loop simulations while requiring 6x less compute. arxiv.org/abs/2506.04218

🚀🚀🚀Announcing our $13M funding round to build the next generation of AI: 𝐒𝐩𝐚𝐭𝐢𝐚𝐥 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥𝐬 that can generate entire 3D environments anchored in space & time. 🚀🚀🚀 Interested? Join our world-class team: 🌍 spaitial.ai youtu.be/FiGX82RUz8U

"ILM "artists" are now being paid to make shimpanzini bananini and bombardiro crocodilo"

Can you train a model for pose estimation directly on casual videos without supervision? Turns out you can! In our #CVPR2025 paper AnyCam, we directly train on YouTube videos and achieve SOTA results by using an uncertainty-based flow loss and monocular priors! ⬇️

New Paper: Continuous Thought Machines pub.sakana.ai/ctm/ Neurons in brains use timing and synchronization in the way that they compute, but this is largely ignored in modern neural nets. We believe neural timing is key for the flexibility and adaptability of biological intelligence. Thread ↓

📣 Excited to share our #CVPR2025 Spotlight paper and my internship project @wayve: SimLingo. A Vision-Language-Action (VLA) model that achieves state-of-the-art driving performance with language capabilities. Code: github.com/RenzKa/simli... Paper: arxiv.org/abs/2503.09594

📢 New paper CVPR 25! Can meshes capture fuzzy geometry? Volumetric Surfaces uses adaptive textured shells to model hair, fur without the splatting / volume overhead. It’s fast, looks great, and runs in real time even on budget phones. 🔗 autonomousvision.github.io/volsurfs/ 📄 arxiv.org/pdf/2409.02482

"Science is an investment. We will put forward a new 500 million package for 2025-2027 to support the best and the brightest researchers and scientists from Europe and around the world." — President @vonderleyen.ec.europa.eu at the ‘Choose Europe for Science' event at La Sorbonne 🇫🇷

Introducing CaRL: Learning Scalable Planning Policies with Simple Rewards We show how simple rewards enable scaling up PPO for planning. CaRL outperforms all prior learning-based approaches on nuPlan Val14 and CARLA longest6 v2, using less inference compute. arxiv.org/abs/2504.17838

Sharing another video showing how LoftUp significantly improves DINOv2 features! Works like a charm! Try it out: Code: github.com/andrehuang/l... Paper: arxiv.org/abs/2504.14032

New CVPR paper by @s-esposito.bsky.social in collaboration with Peter Kontschieder's team at Meta.

Can we represent fuzzy geometry with meshes? "Volumetric Surfaces" uses layered meshes to represent the look of hair, fur & more without the splatting/volume overhead. Fast, pretty, and runs in real-time on your laptop! 🔗 autonomousvision.github.io/volsurfs/ 📄 arxiv.org/pdf/2409.02482

A thread of thoughts on radiance fields, from my keynote at 3DV: Radiance fields have had 3 distinct generations. First was NeRF: just posenc and a tiny MLP. This was slow to train but worked really well, and it was unusually compressed --- The NeRF was smaller than the images.

I remember seeing this drone video a few years ago and thinking "we'll never run SLAM on that".... but here it is, complete with dense reconstruction (single camera, unknown calibration, no IMU). MASt3R-SLAM is absurdly robust.

Apparently, OpenAI is already aligning its LLM to the expectations of the new fascist government – racism, lies, and conspiracy theories will now be sold as «multiple perspectives on controversial subjects» techcrunch.com/2025/02/16/o...

Trying out monocular depth estimation models on abstract images to see what priors they learned. A human might interpret the top as sky and the bottom as a ground, thus giving the top a constant large (blue) depth and the ground a vertical gradient from low (red) to large depths. The model doesn't.

film still #32 Goats #severance photo credit: Trudy Buck

It's ok to use chatbots for everyday monotony like writing your wedding vows because this frees you up to use your creative energy for what truly matters: making powerpoints for work meetings

checking in on deepseek

As someone who has reported on AI for 7 years and covered China tech as well, I think the biggest lesson to be drawn from DeepSeek is the huge cracks it illustrates with the current dominant paradigm of AI development. A long thread. 1/

This week we had our winter retreat jointly with Daniel Cremer's group in Montafon, Austria. 46 talks, 100 Km of slopes and night sledding with some occasionally lost and found. It has been fun!

We propose MET3R, a new metric for measuring multi-view consistency in generated images. Our method is built upon DUSt3R and evaluates the consistency of projected DINO features between two views. It is able to accurately capture the 3D consistency in generated images.

Excited to share that today our paper recommender platform www.scholar-inbox.com has reached 20k users! We hope to reach 100k by the end of the year.. Lots of new features are being worked on currently and rolled out soon.

📢Prepend “Singular” to “Task Vectors” and get +15% average accuracy for free! 1. Perform a low-rank approximation of layer-wise task vectors. 2. Minimize task interference by orthogonalizing inter-task singular vectors. 🧵(1/6)

Instead of listing my publications, as the year draws to an end, I want to shine the spotlight on the commonplace assumption that productivity must always increase. Good research is disruptive and thinking time is central to high quality scholarship and necessary for disruptive research.

A short list of tips for keeping a clean, organized ML codebase for new researchers: eugenevinitsky.com/posts/quick-...

Computer Vision: Fact & Fiction is now available on YouTube 🙌🏼 I made a playlist for it with the seven chapters. Enjoy this time capsule from two decades ago!

Introducing CARLA Garage, a starter kit for developing algorithms for the challenging new CARLA Leaderboard 2.0! Everything you need to step in to autonomous driving research, open-sourced: expert driver, dataset, pretrained models, evaluation, and training scripts. 🔗 github.com/autonomousvi...

Stefano Esposito, Anpei Chen, Christian Reiser, Samuel Rota Bul\`o, Lorenzo Porzi, Katja Schwarz, Christian Richardt, Michael Zollh\"ofer, Peter Kontschieder, Andreas Geiger Volumetric Surfaces: Representing Fuzzy Geometries with Multiple Meshes https://arxiv.org/abs/2409.02482