Unofficial PyTorch implementation of the paper: "CenterNet3D: An Anchor free Object Detector for Autonomous Driving"
-
Updated
Aug 20, 2020 - Python
Unofficial PyTorch implementation of the paper: "CenterNet3D: An Anchor free Object Detector for Autonomous Driving"
50%+ Faster Cylinder3D, compatible with current PyTorch, CUDA, and Spconv versions
spconv-Triton is a fast sparse convolution library with full operator support (SubM conv, Conv3D, Transposed Conv, Pooling). It is hardware-agnostic and runs on GPUs supporting Triton.
Sparse ConvLSTM for Point Cloud Semantic Segmentation
Sparse Autoencoder + Pruning for CMS jet classification | GSoC 2026 @ CERN-HSF | Built with spconv + PyTorch
To associate your repository with the spconv topic, visit your repo's landing page and select "manage topics."