search

mit-han-lab / bevfusion

[ICRA'23] BEVFusion: Multi-Task Multi-Sensor Fusion with Unified Bird's-Eye View Representation
https://bevfusion.mit.edu
Apache License 2.0
2.22k stars 405 forks source link

the MAP is very low, #531

Closed mzy991215 closed 1 month ago

mzy991215 commented 10 months ago

I use your config(lidar-only), and I got the 0 MAP. But when i use the .pth you provided(lidar-only-det.pth) to eval, I can reporduce the standard result. and then I thought there may be a bug in dataset pipeline, so I banned the ObjectPaste，but I still got the wrong mAP about 0.249

Could you help me, thank you!

mAP: 0.0459
mATE: 0.5196 mASE: 0.4398 mAOE: 0.9982 mAVE: 1.0748 mAAE: 0.4462 NDS: 0.1826 Eval time: 101.5s Per-class results: Object Class AP ATE ASE AOE AVE AAE
car 0.068 0.356 0.462 0.695 1.220 0.428 truck 0.015 0.593 0.376 0.585 0.748 0.401 bus 0.061 0.685 0.301 0.893 1.805 0.614 trailer 0.000 0.662 0.328 1.251 0.473 0.346 construction_vehicle 0.000 0.910 0.508 1.550 0.153 0.352 pedestrian 0.226 0.357 0.483 1.144 0.996 0.799 motorcycle 0.000 0.415 0.463 1.189 2.566 0.461 bicycle 0.000 0.418 0.556 1.425 0.638 0.168 traffic_cone 0.013 0.296 0.505 nan nan nan
barrier 0.075 0.505 0.416 0.253 nan nan

Li-Whasaka commented 7 months ago

Have you found problem?

SivenCapo commented 5 months ago

Have you found problem? I encountered the same issue for 1 month. Could we have a Communication?

zhijian-liu commented 1 month ago

Thank you for your interest in our project. This repository is no longer actively maintained, so we will be closing this issue. Please refer to the amazing implementation at MMDetection3D. Thank you again!