ScanNet/ScanNet resource

Health v2 · maintenance only

Flags: no_releases no_license

Adoption not part of the score

2331 stars · 374 forks observed · 2026-08-28

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

ScanNet is a large RGB-D video dataset with 2.5 million views across 1500+ indoor scans, annotated with 3D camera poses, surface reconstructions, and instance-level semantic segmentations. The repository also includes a C++/Python toolkit for parsing the sensor streams and data files.

Use cases

• train 3d scene understanding models on rgb-d data
• semantic segmentation of indoor 3d scans
• benchmark 3d object detection and voxel labeling
• evaluate rgb-d slam and camera pose estimation
• get annotated 3d meshes of indoor scenes for deep learning
• cad model retrieval research dataset

When to choose

• you need large-scale annotated RGB-D indoor scene data for 3D vision research
• you want a standard benchmark for 3D semantic segmentation or object detection
• you need camera poses plus reconstructed meshes together

When to avoid

• you need outdoor or LiDAR data
• you cannot agree to the ScanNet Terms of Use or use a non-institutional email
• you need a ready-to-use model rather than a dataset

Facets

dataset · maturity stable

computer-vision machine-learning image-processing data-science computer-vision deep-learning graphics artificial-intelligence cpp python cross-platform rgbd 3d-reconstruction semantic-segmentation benchmark indoor-scenes 3d-scene-understanding meshes camera-pose linux

Member repositories

For agents

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