Haiyang-W / RBGNet

[CVPR2022] This is the official code of "RBGNet: Ray-based Grouping for 3D Object Detection".
41 stars 4 forks source link

recall_0.25=(0, 0) / 0 #5

Closed LIUXIN4558 closed 3 months ago

LIUXIN4558 commented 3 months ago
Hello, may I ask why I kept recall_0.25 at 0 while running the SUNRGBD validation set:
eval: 100%|███████████████████████████████████████████████████████████| 5050/5050 [25:24<00:00,  3.31it/s, recall_0.25=(0, 0) / 0]
2024-07-28 08:33:17,466   INFO  *************** Performance of EPOCH 8 *****************
2024-07-28 08:33:17,466   INFO  Generate label finished(sec_per_example: 0.3020 second).
2024-07-28 08:33:17,467   INFO  recall_roi_0.25: 0.000000
2024-07-28 08:33:17,467   INFO  recall_rcnn_0.25: 0.000000
2024-07-28 08:33:17,467   INFO  recall_roi_0.5: 0.000000
2024-07-28 08:33:17,467   INFO  recall_rcnn_0.5: 0.000000

Looking forward to your reply

Haiyang-W commented 3 months ago

The recall here is not implemented, so it doesn't need to be referenced. Are there any issues with the normal test results?

LIUXIN4558 commented 3 months ago

Thank you very much for your reply. There are currently no further issues!

Haiyang-W commented 3 months ago

I recommend you use CAGroup3D. Its performance is higher, and the code base is cleaner.