We introduce an LRM capable of recovering illumination, geometry and material properties of real object scenes from a few posed images.

We propose a method for compositional part-level 3D generation and reconstruction from various modalities including text, image or 3D models.

We combine Meta 3D AssetGen and TextureGen for high-quality mesh generation.

We propose a method that encodes 2D images into any 3D representation, without requiring pre-trained image feature extractor.

We build upon sprite-based image decomposition approaches to design a generative method for character analysis and recognition in text lines.

We introduce MACARONS, a method that learns in a self-supervised fashion to explore new environments and reconstruct them in 3D using RGB images only.

A Transformer-based framework to evaluate off-the-shelf features (object-centric and dense representations) for the reasoning task of VQA.

We characterize 3D shapes as affine transformations of linear families learned without supervision, and showcase its advantages on large shape collections.

Leveraging synthetic training data to efficiently extract visual elements from historical document images.