minkowski-engine

Minkowski Engine is an auto-diff neural network library for high-dimensional sparse tensors

4D Spatio-Temporal Semantic Segmentation on a 3D video (a sequence of 3D scans)

Building Volumetric Beliefs for Dynamic Environments Exploiting Map-Based Moving Object Segmentation (RAL 2023)

MinkLoc3D: Point Cloud Based Large-Scale Place Recognition

MinkLoc++: Lidar and Monocular Image Fusion for Place Recognition

MinkLoc3Dv2: Improving Point Cloud Based Place Recognition with Ranking-based Loss and Large Batch Training

EgoNN: Egocentric Neural Network for Point Cloud Based 6DoF Relocalization at the City Scale

MINSU3D: MinkowskiEngine-powered Scene Understanding in 3D

Generalizable Stable Points Segmentation for 3D LiDAR Scan-to-Map Long-Term Localization

Project examing sparse deep learning architectures for ligand classification.

Fork of NVIDIA MinkowskiEngine modernized for CUDA 12.8+/Blackwell (RTX 50-series), NumPy 2.0, PyTorch 2.x prebuilt wheels, fp16/bf16 autocast, fused gather/scatter. Best-effort, no support commitment.

3D Reconstruction and Classification from Very High Resolution Satellite Imagery (ReKlaSat 3D)

NVIDIA/MinkowskiEngine ported to torch 2.11 / CUDA 12.8 / CCCL 2.x: builds and runs on sm_89 + sm_120 (Blackwell), 48 errors to 0

A toolkit for preprocessing, training and inference