🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

The importance of stupidity in scientific research!

We were awarded an ERC Starting Grant #ERCStG to support Mathieu Aubry's project on discovering and analyzing visual structures (DISCOVER), with applications for History and Earth imagery ! European Research Council (ERC)

[ISMIR 2022] "A Model You Can Hear: Audio Identification with Playable Prototypes" Revisiting transformation-invariant clustering for audio with SOTA results! Check out the paper arxiv.org/abs/2208.03311 #ISMIR2022 #MachineLearning #Interpretable #Supervised #Unsupervised

Data drives progress in computer vision. We introduce Infinigen: a generator of unlimited high-quality 3D data. 100% procedural, no external assets, no AI. Free and open source. Intro video: youtu.be/6tgspeI-GHY Code: github.com/princeton-vl/i… CVPR23: Wed 4:30PM, Poster 27

We release Llemma: open LMs for math trained on up to 200B tokens of mathematical text. The performance of Llemma 34B approaches Google's Minerva 62B despite having half the parameters. Models/data/code: github.com/EleutherAI/mat… Paper: arxiv.org/abs/2310.10631 More ⬇️

🚨The FLAIR dataset has been accepted at NeurIPS2023 in the Benchmark & Dataset track. It contains high-resolution aerial imagery with a ground sample distance of 20 cm and over 20 billion individually labeled pixels for precise land-cover classification. arxiv.org/abs/2310.13336

SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Rendering paper page: huggingface.co/papers/2311.12… propose a method to allow precise and extremely fast mesh extraction from 3D Gaussian Splatting. Gaussian Splatting has recently

Infinigen v1.1.1 is out: updated to Blender3.6 and revamped installation to be more robust. See github.com/princeton-vl/i… for more, including our "minimal" install variant which lets you generate non-terrain assets in almost any python environment (no local compilation required).

Announcing the 1st Workshop on Dataset Distillation #CVPR2025 2024! #CVPR2024 sites.google.com/view/dd-cvpr20… More info in 🧵 1/4

Introducing “FlowMap”, the first self-supervised, differentiable structure-from-motion method that is competitive with conventional SfM like Colmap! cameronosmith.github.io/flowmap/ IMO this solves a major missing piece for internet-scale training of 3D Deep Learning methods. 1/n

Amazing work from the group, and lots of directions opened by differentiable SfM! On top of this, the codebase is so neat and sweet to use Extremely exciting to be here at such times!

Announcing (fashionably late) our #CVPR2024 paper on GenAI detection! with Avneesh Sud, Thomas Leung, and Ben Usman Fake-Inversion: Learning to Detect Images from Unseen Models by Inverting Stable Diffusion fake-inversion.github.io More info in 🧵 1/7

Introducing Neural Isometries where we show how to exploit equivariant ML even for transformations that are “nasty”, e.g. non-compact, projective, nonlinear, or not even a group action! arxiv.org/abs/2405.19296 Collab w/ the amazing Tommy Mitchel Tommy Mitchel and Mike Taylor! 1/n

Schopenhauer, the art of not reading

Introducing Diffusion Forcing, which unifies next-token prediction (eg LLMs) and full-seq. diffusion (eg SORA)! It offers improved performance & new sampling strategies in vision and robotics, such as stable, infinite video generation, better diffusion planning, and more! (1/8)

The Learnable Typewriter A Generative Approach to Text Analysis w/ Nicolas Gonthier Julien Gaubil Tom Monnier and Mathieu Aubry ⭐️ won the best paper award at #ICDAR24 learnable-typewriter.github.io

Superb news - all thanks to Yannis Siglidis amazing work, it was a blast to collaborate on this project!

💻We've released the code for our #CVPR2025 paper MAtCha! 🍵MAtCha reconstructs sharp, accurate and scalable meshes of both foreground AND background from just a few unposed images (eg 3 to 10 images)... ...While also working with dense-view datasets (hundreds of images)!