nianticlabs/simplerecon

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

1428 stars · 130 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

SimpleRecon is the reference PyTorch implementation of an ECCV 2022 paper for multi-view stereo depth estimation and 3D reconstruction that avoids 3D convolutions. It provides training, evaluation, point cloud fusion, mesh metrics, and visualization tooling for MVS depth models, primarily benchmarked on ScanNetv2.

Use cases

• estimate depth maps from multiple posed images
• reconstruct 3D point clouds or meshes from video frames
• evaluate multi-view stereo models on ScanNet
• train or finetune an MVS depth estimation network
• visualize cost volumes and depth predictions
• run 3D scene reconstruction without 3D convolutions

When to choose

• you need state-of-the-art multi-view stereo depth estimation with a fast 2D-CNN architecture
• you want a research baseline or pretrained weights for MVS on ScanNetv2
• you need point cloud fusion and mesh evaluation tooling alongside depth prediction

When to avoid

• you need a commercially licensed model - the code is non-commercial use only
• you want a production-ready 3D reconstruction pipeline rather than research code
• you lack posed multi-view images with camera intrinsics
• you need real-time reconstruction on CPU or mobile hardware

Facets

library · maturity maintenance

machine-learning computer-vision image-processing data-visualization computer-vision deep-learning machine-learning python multi-view-stereo depth-estimation cost-volume 3d-reconstruction pytorch eccv2022 scannet point-cloud-fusion research-code non-commercial-license gpu linux

Member repositories

For agents

markdown · JSON · MCP: product_card(name="nianticlabs/simplerecon")