DeepInteraction: 3D Object Detection via Modality Interaction,
Zeyu Yang, Jiaqi Chen, Zhenwei Miao, Wei Li, Xiatian Zhu, Li Zhang
NeurIPS 2022
- (2022/6/27) DeepInteraction-e ranks first on nuScenes among all solutions.
- (2022/6/26) DeepInteraction-large ranks first on nuScenes among all non-ensemble solutions.
- (2022/5/18) DeepInteraction-base ranks first on nuScenes among all solutions that do not use test-time augmentation and model ensemble.
Model | Modality | mAP | NDS |
---|---|---|---|
DeepInteraction-e | C+L | 75.74 | 76.34 |
DeepInteraction-large | C+L | 74.12 | 75.52 |
DeepInteraction-base | C+L | 70.78 | 73.43 |
Model | Modality | mAP | NDS | Checkpoint |
---|---|---|---|---|
DeepInteraction-base | C+L | 69.85 | 72.63 | Fusion_0075_refactor.pth |
This implementation is build upon mmdetection3d, please follow the steps in install.md to prepare the environment.
Please follow the official instructions of mmdetection3d to process the nuScenes dataset.(https://mmdetection3d.readthedocs.io/en/latest/datasets/nuscenes_det.html)
Downloads the pretrained backbone weights to pretrained/
# train with 8 GPUs
tools/dist_train.sh projects/configs/nuscenes/Fusion_0075_refactor.py 8
# test with 8 GPUs
tools/dist_test.sh projects/configs/nuscenes/Fusion_0075_refactor.py ${CHECKPOINT_FILE} 8 --eval=bbox
Many thanks to the following open-source projects:
@inproceedings{yang2022deepinteraction,
title={DeepInteraction: 3D Object Detection via Modality Interaction},
author={Yang, Zeyu and Chen, Jiaqi and Miao, Zhenwei and Li, Wei and Zhu, Xiatian and Zhang, Li},
booktitle={NeurIPS},
year={2022}
}