readme.md

June 26, 2023 ยท View on GitHub

BEVDistill

Cross-Modal BEV Distillation for Multi-View 3D Object Detection


Zehui Chen, Zhenyu Li, Shiquan Zhang, Liangji Fang, Qinhong Jiang, Feng Zhao.

NOTE

This code is an initial release with BEVFormer-T only, we will release a full version in the future.

Run

# baseline
./tools/dist_train.sh projects/configs/video_bevformer/bevformer_128x128_r50_2x.py 8
# BEVDistill
./tools/dist_train_distill.sh projects/configs/bevdistill/bevdistill_r50_128x128_900query_bboxrweighttop100_bevgt_mocov2_merge_r1_2x.py 8

Performance

nuScenes Val set

ModelconfigmAPNDS
BEVFormer-R5035.242.3
BEVDistill-R5038.645.7
BEVFormer-T-R50link38.348.8
BEVDistill-T-R50link40.751.5
BEVFormer-T-R101link40.550.6
BEVDistill-T-R101link41.752.4

Pretrain Checkpoints

ModelWeightmAPNDS
ObjectDGCNN (Teacher)link67.461.5
FCOS3D r50 (Student)link