This repository is a curated list of papers, code, and resources related to DUSt3R and MASt3R, which are geometric foundation models for 3D computer vision tasks. It serves researchers and practitioners in the field of 3D reconstruction, pose estimation, and scene understanding, providing a centralized hub for the latest advancements and tools.

How It Works

DUSt3R and its successors represent a paradigm shift in 3D vision by framing tasks like multi-view stereo reconstruction and image matching as regression problems. Instead of relying on traditional, often brittle, pipelines involving explicit camera calibration and feature matching, these models leverage Transformer architectures to directly predict "pointmaps" (pixel-aligned 3D point clouds). This approach allows for unconstrained reconstruction from arbitrary image collections, even without prior knowledge of camera poses or intrinsics, and can unify various 3D vision tasks.

Quick Start & Requirements

This repository is a collection of links and does not have a direct installation or run command. Users are directed to individual project pages for specific codebases and their requirements.

Highlighted Details

• Foundation Model: DUSt3R and MASt3R are presented as foundational models that simplify and unify multiple 3D vision tasks.
• Unconstrained Reconstruction: The core models operate without requiring prior camera calibration or pose information.
• Extensive Coverage: The list tracks a wide array of related works, including Gaussian Splatting, Structure-from-Motion, dynamic scene reconstruction, scene understanding, and robotics applications.
• Regular Updates: The repository is actively maintained, with frequent additions of new papers and codebases.

Maintenance & Community

The repository is maintained by Rui Li. It references other "awesome" lists for 3D vision, indicating community engagement and a desire to build upon existing resources.

Licensing & Compatibility

The repository itself is a list and does not have a license. Individual linked projects will have their own licenses, which users must consult for compatibility and usage restrictions.

Limitations & Caveats

As a curated list, this repository does not provide direct functionality. Users must navigate to external links for code and resources, and the effectiveness of the underlying models depends on their specific implementations and training data.