JonasSchult/Mask3D
Mask3D predicts accurate 3D semantic instances achieving state-of-the-art on ScanNet, ScanNet200, S3DIS and STPLS3D.
Builds on MinkowskiEngine sparse convolutions and transformer-based mask prediction to decouple semantic and instance segmentation into separate decoder branches, eliminating hand-crafted grouping heuristics. Implemented in PyTorch with PyTorch Lightning for training orchestration and Hydra for modular configuration management. Supports indoor 3D scene understanding across multiple datasets with preprocessing pipelines for ScanNet, S3DIS, and STPLS3D point clouds.
716 stars. No commits in the last 6 months.
Stars
716
Forks
126
Language
Python
License
MIT
Category
Last pushed
Oct 29, 2023
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/computer-vision/JonasSchult/Mask3D"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Related tools
nsavinov/semantic3dnet
Point cloud semantic segmentation via Deep 3D Convolutional Neural Network
PRBonn/LiDAR-MOS
(LMNet) Moving Object Segmentation in 3D LiDAR Data: A Learning-based Approach Exploiting...
yanx27/2DPASS
2DPASS: 2D Priors Assisted Semantic Segmentation on LiDAR Point Clouds (ECCV 2022) :fire:
ctu-vras/traversability_estimation
Semantic Segmentation of Images and Point Clouds for Traversability Estimation
suyogduttjain/fusionseg
Video Object Segmentation