New generation of CO3D data, now from 1000+ object categories, with gaussian splats reconstruction!

🍺 LagerNVS (CVPR 2026) 🍺 LagerNVS is a generalizable, feed-forward, real-time Novel View Synthesis network which - performs rendering in real time, - generalizes to in-the-wild data, - works with and without known source cameras, - sets a new state-of-the-art among deterministic methods, - can be paired with a diffusion decoder for generative extrapolation. LagerNVS shows that 3D biases are useful for Novel View Synthesis but explicit 3D representations are not required to achieve them. We use 3D biases in (1) architecture design and (2) pre-training: (1) In NVS with explicit 3D representations (3DGS, NeRF) reconstruction is typically difficult and slow, but rendering is much faster and simpler. We mimic this process in the network design: we use a large (1B params) encoder and a small, lightweight decoder (ViT-B). This allows increasing the network capacity while still achieving real-time rendering. (2) The encoder, initialized from VGGT, was pre-trained with 3D reconstruction objectives, making the initial features 3D aware. Both substantially improve performance. Project page: szymanowiczs.github.io/lager… Code: github.com/facebookresearch/… Paper: arxiv.org/abs/2603.20176 Models: huggingface.co/collections/f… Work done with @jianyuan_wang @MinghaoChen23 Christian Rupprecht and Andrea Vedaldi

Excited to introduce 🎬ActionMesh, a fast model transforming any video → high-quality animated 3D mesh ! Generate animated mesh seamlessly importable into any 3D software in less than a minute. 🤗Try it out: huggingface.co/spaces/facebo… 🌐Project Page: remysabathier.github.io/acti… 📄Paper: remysabathier.github.io/acti… 💻Code: github.com/facebookresearch/… #Video4D #GenAI #3DGeneration @Meta @RealityLabs @AIatMeta @ucl

Our team moved from object generation to generating the whole navigable scenes! Tech report: scontent-lhr6-2.xx.fbcdn.net…

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/facebookresearch/… 📄Paper: arxiv.org/abs/2501.07574

Our latest improvement to CO3D: more objects, more categories, more modalities! uco3d.github.io/

🥳Excited to share my recent work at Meta, "PartGen: Part-level 3D Generation and Reconstruction with Multi-View Diffusion Models", which aims at compositional/part-level 3D generation and reconstruction from various modalities. Project page: silent-chen.github.io/PartGe…

We’re also publishing two research papers on the approaches we used for high quality 3D model generation & texture generation based on text prompts that are behind the Meta 3D Gen system. Meta 3D AssetGen ➡️ go.fb.me/n87zde Meta 3D TextureGen ➡️ go.fb.me/gngy5i

📣 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, producing results that are superior to existing solutions — at 3-10x the speed of existing work in this space. Details in the technical report ➡️ go.fb.me/ry96ln

Excited to release Lightplane: A highly memory-efficient differentiable NeRF renderer and feature splatter. It renders FullHD image batches using <1GB of GPU memory and greatly scales up 3D deep models such as the Large Reconstruction Model. With @AngCao3, @jcjohnss, Andrea Vedaldi Project page: lightplane.github.io/ Arxiv: arxiv.org/abs/2404.19760 Github: github.com/facebookresearch/… Docs: lightplane.github.io/docs/

Do you like cats and dogs and always wanted a large-scale video dataset for 3D reconstruction??? 🐈 🐶 Join me and @davnov134 as we present our ✨highlight✨ poster at 4:30 PM @CVPR #CVPR2023 Website: cop3d.github.io Arxiv: arxiv.org/abs/2211.03889

Novel-view synthesis has achived great success, but can we also synthesize the sound at novel viewpoints? We present "Novel-view Acoustic Synthesis" @CVPR Tues PM. We introduce the very first NVAS formulation, multi-view audio-visual datasets, and approach to this problem.

RT @zooko: 🤔 Milton Friedman's thermostat worthwhile.typepad.com/worth…

I put some famous logos through ControlNet.

It's been almost a year since I started a math club with a couple of children from my son's class. This is the thread about how I run it, a list of recommended resources and some other related stuff. 🧵1/n

There are many reflections than can be made over what this map shows. Interesting.

We are hiring PhD student interns to work on 3D/4D reconstruction / conditional generation! Feel free to get in touch if you want to apply or get more information. metacareers.com/jobs/7211456…

Introducing Make-A-Video3D! Generating 3D dynamic (mini) scenes from input text. That is, text --> 4D! Needs no 4D data (i.e., no dynamic 3D data), no static 3D data, no paired text-video data. Paper: arxiv.org/abs/2301.11280 Website: make-a-video3d.github.io/

This is a great programme for getting into AI research after undergrad or switching from another field! Ask @_sam_sinha_ . :)

The future of email with ChatGPT.