Released code for the paper: a single-stage monocular 3D object detection model is proposed. An instance-segmentation head is integrated into the model training, which allows the model to be aware of the visible shape of a target object.
The code is tested with
python==3.7, torch==1.4.0, torchvision==0.5.0
.
After python and pytorch are installed on your machine, You can run the following commands to prepare the environment:
pip install -r requirements.txt
cd mmdet3d/ops/DCNv2
. make.sh
cd ../../..
pip install -v -e .
Please download the KITTI dataset and organize the data as follows
data/kitti/
|training/
|calib/
|image_2/
|label/
|ImageSets/
|testing/
|calib/
|image_2/
|ImageSets/
training:
python -m torch.distributed.launch --nproc_per_node=<number of gpus> mono3d/train.py <config file> --no-validate --launcher=pytorch
evaluation:
python mono3d/test.py <config file> <checkpoint file> --eval bbox
Config File | Model | AP(3D) |
---|---|---|
configs_shapeaware/shapeaware_quantizationFloor_albu_SAFLw2NoRepair_CS.py | Google Drive | 18.39 |
The project benefits a lot from the following works. Thanks for their contribution.