We provide the experiments on the NuScenes dataset. The results are as follows:
Camera | mAP | mATE | mASE | mAOE | mAVE | mAAE | NDS |
---|---|---|---|---|---|---|---|
front | 1.670 | 0.882 | 0.699 | 0.720 | 1.506 | 1.0 | 7.814 |
front_left | 0.066 | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | 0.033 |
front_right | 0.138 | 0.956 | 0.934 | 1.0 | 1.0 | 1.0 | 1.163 |
back | 2.683 | 0.835 | 0.701 | 0.721 | 1.300 | 1.0 | 8.767 |
back_left | 0.040 | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | 0.020 |
back_right | 0.138 | 1.0 | 1.0 | 1.0 | 1.0 | 1.0 | 0.069 |
all | 18.220 | 0.717 | 0.414 | 0.826 | 1.452 | 1.0 | 19.540 |
To run our code on NuScenes, one can run
python scripts/nuscenes2kitti.py
to convert to NuScenes dataset to KITTI format. Then change the path of KITTI to NuScenes and kick off the training.