Code for G3T and G3T-Long

A PyTorch library for implementing flow matching algorithms, featuring continuous and discrete flow matching implementations. It includes practical examples for both text and image modalities.

Official implement of VGGT-Long

Any4D: Unified Feed-Forward Metric 4D Reconstruction

Official code for paper "InstantSfM: Towards GPU-Native SfM for the Deep Learning Era"

[ICLR 2026] π^3: Permutation-Equivariant Visual Geometry Learning

✨ An advanced 3D Gaussian Splatting renderer for THREE.js

[CVPR 2025] "DiffusionSfM: Predicting Structure and Motion via Ray Origin and Endpoint Diffusion" official implementation.

MP-SfM: Monocular Surface Priors for Robust Structure-from-Motion (CVPR 2025)

[ICCV 2025 Highlight] Geo4D: Leveraging Video Generators for Geometric 4D Scene Reconstruction

[NeurIPS 2024] Official code for "Neural Gaffer: Relighting Any Object via Diffusion"

[ICLR 2025 Oral] Official code for "LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias"

[CVPR 2025 Best Paper Award] VGGT: Visual Geometry Grounded Transformer

Code for the project "MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos"

[CVPR 2024] RoMa: Robust Dense Feature Matching; RoMa is the robust dense feature matcher capable of estimating pixel-dense warps and reliable certainties for almost any image pair.

Minimal solvers for calibrated camera pose estimation

Official implementation of Continuous 3D Perception Model with Persistent State

Official repo for paper "Structured 3D Latents for Scalable and Versatile 3D Generation" (CVPR'25 Spotlight).

Code for "Accurate Differential Operators for Hybrid Neural Fields", accepted at CVPR 2025

NVIDIA Cosmos is an open platform of world models, datasets, and tools that enables developers to build Physical AI for robots, autonomous vehicles, smart infrastructure, and more.

A toolbox for mapping and localization with line features.

Lightplane implements a highly memory-efficient differentiable radiance field renderer, and a module for unprojecting features from images to 3D grids.

Web-based 3D visualization in Python

A library for differentiable robotics on manifolds.

Simulation platform for general-purpose robotics & embodied AI learning.

[CVPR 2023 Highlight] Perspective Fields for Single Image Camera Calibration

(Deprecated) Python bindings for COLMAP

[CVPR'25]Tora: Trajectory-oriented Diffusion Transformer for Video Generation