Reproduce “Learning a Multi-View Stereo Machine” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Abhishek Kar, Christian Häne, Jitendra Malik

V-LSM reconstructs 3D object shape from posed images by unprojecting image features into a shared metric grid, recurrently fusing views, and predicting voxel occupancy. On synthetic ShapeNet, with four views, it achieves mean voxel IoU of 78.2, versus 62.1 for the authors’ pose-aware 3D-R2N2 reimplementation. The result is limited to synthetic objects, known poses, and a 32³ grid.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.