Implementation of 'FADet: A Multi-sensor 3D Object Detection Network based on Local Featured Attention'
Firstly,
git clone https://github.com/ZionGo6/FADet
We recommend to run our FADet in docker container, dockerfile and dependencies are prepared in this repo.
cd FADet/docker/
To build docker image, run:
./01_build_image.sh
Change the parameters inside as you need then run the container:
./02_run_container.sh
In the docker container, run:
pip install -v -e .
We use full nuScenes dataset, please download and process dataset based on here. Finally, the data/ should be like this:
├── data
│ ├── nuscenes
│ │ ├── maps
│ │ ├── samples
│ │ ├── sweeps
│ │ ├── v1.0-test
| | ├── v1.0-trainval
│ │ ├── nuscenes_database
│ │ ├── nuscenes_infos_train.pkl
│ │ ├── nuscenes_infos_val.pkl
│ │ ├── nuscenes_infos_test.pkl
│ │ ├── nuscenes_dbinfos_train.pkl
Will release soon