awesome-rocm-autodrive

A curated and practical collection of training examples, tools, and ROCm optimizations for key autonomous driving models. It aims to accelerate the deployment of deep learning models on AMD GPUs by providing out-of-the-box training support, Docker environments, and performance tuning utilities.

🔧 Supported Models (Initial Release)

Model	Type	Repo Link	README for ROCm	Notes
ResNet-50	Backbone	https://github.com/amd-fuweiy/vision	example/backbone
EfficientNet-B7	Backbone	https://github.com/amd-fuweiy/vision		Currently have performance issue with DWConv
PointPillars	Point Cloud	https://github.com/Treemann/mmdetection3d	examples/detection/pointpillars	Need ROCm mmcv to get better performance
MapTR	Vector Prediction	https://github.com/aaab8b/MapTR	See modified Readme in git
FlashOcc	Scene Occupancy	https://github.com/mingjielu/FlashOCC	examples/occupancy/FlashOCC
Sparse4D	Sparse Detector	https://github.com/binding7012/Sparse4D	examples/detection/sparse4d
BEVFormer	Multi-view	https://github.com/jun-amd/BEVFormer	examples/detection/BEVFormer
PETR	3D Detection	https://github.com/aaab8b/PETR	See modified Readme in git
QCNet	Trajectory Prediction	https://github.com/aaab8b/QCNet	See modified Readme in git
SurroundOcc	Scene Occupancy	https://github.com/mingjielu/SurroundOcc	examples/occupancy/FlashOCC
UniAD	Multi-model	https://github.com/mingjielu/UniAD	examples/Multi-model/UniAD

---

🚀 Quick Start

1. Clone the repo:

git clone https://github.com/YOUR_ORG/awesome-rocm-autodrive.git
cd awesome-rocm-autodrive