Drivable 3D Gaussian Avatars paper page: huggingface.co/papers/2311.08… present Drivable 3D Gaussian Avatars (D3GA), the first 3D controllable model for human bodies rendered with Gaussian splats. Current photorealistic drivable avatars require either accurate 3D registrations during

Looking forward to the next seminar by Michael Zollhoefer! Michael and his team at Reality Labs Research have developed insanely cool technology on Codec Telepresence! Students at UMD Department of Computer Science UMD Center for Machine Learning UMIACS, don't miss the talk!

The Codec Avatars Lab in Pittsburgh is looking for summer Research Scientist interns. Reach out to me if you are interested to work on novel neural reconstruction/rendering approaches, digital humans, view synthesis, generative models, or audio research. metacareers.com/jobs/336581275…

HybridNeRF: Efficient Neural Rendering via Adaptive Volumetric Surfaces paper page: huggingface.co/papers/2312.03… Neural radiance fields provide state-of-the-art view synthesis quality but tend to be slow to render. One reason is that they make use of volume rendering, thus requiring

HybridNeRF combines the best of NeRF with SDFs using a hybrid surface–volume representation. 🤩State-of-the-art visual quality 🚀40+ FPS at 2K×2K VR on a single 4090 GPU (almost 10× faster than VR-NeRF!) Project: haithemturki.com/hybrid-nerf/ Paper: arxiv.org/abs/2312.03160

Haithem Turki is also presenting PyNeRF at NeurIPS Conference 2023 next week. Talk to him to find out more! :) x.com/RadianceFields…

(1/3) 𝐕𝐨𝐱𝐞𝐥 𝐇𝐚𝐬𝐡𝐢𝐧𝐠 received the Test-of-Time Award SIGGRAPH Asia ➡️ Hong Kong! What an honor together with Michael Zollhoefer Shahram Izadi @mcstammi Voxel hashing is a sparse & efficient data structure for 3D scenes/grids! What's the core idea and why is still relevant today? ⬇️

SpecNeRF: Gaussian Directional Encoding for Specular Reflections paper page: huggingface.co/papers/2312.13… Neural radiance fields have achieved remarkable performance in modeling the appearance of 3D scenes. However, existing approaches still struggle with the view-dependent

Meta announces URHand Universal Relightable Hands paper page: huggingface.co/papers/2401.05… model is a high-fidelity Universal prior for Relightable Hands built upon light-stage data. It generalizes to novel viewpoints, poses, identities, and illuminations, which enables quick

Nerfstudio 1.0 is out and it includes support for our Eyeful Tower dataset containing 11 high-res, room-scale HDR scenes! 📷unprecedented density: 28,572 photos 🔍unprecedented resolution: up to 50 MP Learn more: github.com/facebookresear…

(1/3) Can we turn text-to-image models into photorealistic 3D generators? ViewDiff (#CVPR2024) produces realistic, multi-view consistent images of real-world 3D objects in authentic surroundings. Website lukashoel.github.io/ViewDiff Video youtu.be/SdjoCqHzMMk How does it work?

It was amazing to have our CTO visit and check out the cutting-edge systems we are developing in Pittsburgh.

The "Edge Gradients" paper was accepted to #ECCV2024 🥳! European Conference on Computer Vision #ECCV2026 "Rasterized Edge Gradients: Handling Discontinuities Differentiably" arxiv.org/abs/2405.02508

Upgrade your differentiable renderers with Stan's cool work based on rasterized edge gradients!

(1/2) How to accelerate the reconstruction of 3D Gaussian Splatting? 3DGS-LM replaces the commonly used ADAM optimizer with a tailored Levenberg-Marquardt (LM). => We are 𝟑𝟎% 𝐟𝐚𝐬𝐭𝐞𝐫 𝐭𝐡𝐚𝐧 𝟑𝐃𝐆𝐒 for the same quality. lukashoel.github.io/3DGS-LM/ youtu.be/tDiGuGMssg8

If you'd like to accelerate 3DGS training, check out a recent work of Lukas Höllein leveraging Levenberg-Marquardt algorithm!

Looking for a research internship in 2025? The Social AI Research group at Meta’s Codec Avatars Lab in Pittsburgh is offering topics such as neural rendering, body tracking, motion synthesis, and animation from multimodal sensor inputs. Link: metacareers.com/jobs/822652889…

I'm excited to present "Fillerbuster: Multi-View Scene Completion for Casual Captures"! This is work with my amazing collaborators Norman Müller, Yash Kant, Vasu Agrawal, Michael Zollhoefer, Angjoo Kanazawa, Christian Richardt during my internship at Meta Reality Labs. ethanweber.me/fillerbuster/

🚀New from Meta FAIR: today we’re introducing Seamless Interaction, a research project dedicated to modeling interpersonal dynamics. The project features a family of audiovisual behavioral models, developed in collaboration with Meta’s Codec Avatars lab + Core AI lab, that

My research group in the Codec Avatars lab at Meta contributed by developing the technology required to display the outputs of FAIR SeamlessNext's dyadic motion models as 3D Full-body Codec Avatars.