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

We have an open internship position on socially aware navigation, human aware world models etc.

At @naverlabseurope.bsky.social in Grenoble (Meylan), France.

careers.werecruit.io/en/naver-lab...

The PAISS summer school is back with an incredible line of speakers (and more to come). Spread the word !

AnySat has been accepted as a ✨ highlight at #CVPR2025! See you in Nashville 🎉

We’ll also be presenting this work at:
📍 @egu.eu on 02/04 in Vienna
📍 @esa.int / NASA Workshop on Foundation Models on 05/04 in Rome

We are happy to share LPOSS: Label Propagation Over Patches and Pixels for Open-vocabulary Semantic Segmentation.

LPOSS is a training-free method for open-vocabulary semantic segmentation using Vision-Language Models.

ILIAS is a large-scale test dataset for evaluation on Instance-Level Image retrieval At Scale. It is designed to support future research in image-to-image and text-to-image retrieval for particular objects and serves as a benchmark for evaluating foundation models and retrieval techniques.

Welcome to NAVER LABS Europe on Bluesky! We chose 'LABS blue' to kickoff our posts & we'd appreciate it if you could share this one widely tinyurl.com/2nt9jdnv 😊 #hiring #jobs #internships

(1/3) Happy to share LUDVIG: Learning-free Uplifting of 2D Visual features to Gaussian Splatting scenes, that uplifts visual features from models such as DINOv2 (left) & CLIP (mid) to 3DGS scenes. Joint work w. @dlarlus.bsky.social @jmairal.bsky.social
Webpage & code: juliettemarrie.github.io/ludvig

Physics IQ Benchmark: Do generative video models learn physical principles from watching videos?

TL;DR: comprehensive benchmark dataset; various physical principles (fluid dynamics, optics, solid mechanics, magnetism and thermodynamics); Performed on recent video generators

Outstanding Finalist 2: “DINOv2: Learning Robust Visual Features without Supervision," by Maxime Oquab, Timothée Darcet, Théo Moutakanni et al. 5/n openreview.net/forum?id=a68...