Our #DTIClustering paper has been accepted at #NeurIPS2020 as an oral presentation! It is state-of-the-art, strongly interpretable and works on images in the wild like Instagram hashtags. Code and details: imagine.enpc.fr/~monniert/DTIC… w/ Thibault GROUEIX

Unsupervised Layered Image Decomposition into Object Prototypes pdf: arxiv.org/pdf/2104.14575… abs: arxiv.org/abs/2104.14575 the first layered image decomposition algorithm that learns an explicit and shared concept of object type, and is robust enough to be applied to real images

#ICCV2021 Session 7 “Unsupervised Layered Image Decomposition Into Object Prototypes” by Tom Monnier, Elliot Vincent, Jean Ponce, Mathieu Aubry. arxiv: arxiv.org/abs/2104.14575 website: imagine.enpc.fr/~monniert/DTI-… github: github.com/monniert/dti-s…

Here is a summary thread for our #3DV2021 papers which will be presented this week International Conference on 3D Vision 👇

Very excited I can finally speak about our new unsupervised SVR project called UNICORN🦄!! More details at imagine.enpc.fr/~monniert/UNIC…

Differentiable Blocks World: Qualitative 3D Decomposition by Rendering Primitives Tom Monnier, Jake Austin, Angjoo Kanazawa , Alexei A. Efros, Mathieu Aubry tl;dr: fit textured superquadratics to multiview images. arxiv.org/abs/2307.05473

📣 New research from GenAI at Meta, introducing Meta 3D Gen: A new system for end-to-end generation of 3D assets from text in <1min. Meta 3D Gen is a new combined AI system that can generate high-quality 3D assets, with both high-resolution textures and material maps end-to-end,

UnCommon Objects in 3D

We are releasing uCO3D! Built to supercharge 3D GenAI and digital-twin models, this evolution of CO3D features more and higher-quality object videos from 1k categories, 3D Gaussian Splats, and streamlined OSS tools. 💻Data&code: github.com/facebookresear… 📄Paper:

💥New preprint! WildCAT3D uses tourist photos in-the-wild as supervision to learn to generate novel, consistent views of scenes like the one shown below. Work from Meta AI internship, with @davnov123 Filippos Kokkinos Hadar Averbuch-Elor Tom Monnier (1/5)

Introducing DINOv3: a state-of-the-art computer vision model trained with self-supervised learning (SSL) that produces powerful, high-resolution image features. For the first time, a single frozen vision backbone outperforms specialized solutions on multiple long-standing dense