search

Czm369 / MixPL

Mixed Pseudo Labels for Semi-Supervised Object Detection
Apache License 2.0
47 stars 0 forks source link

The result I reproduced is almost 0. #9

Closed gaby20000413 closed 1 month ago

gaby20000413 commented 3 months ago

I followed this website to download the COCO dataset and trained the model by splitting the dataset accordingly. config: \projects\MixPL\configs\mixpl_faster-rcnn_r50-caffe_fpn_180k_coco-s1-p10.py train: \tools\train.py hardware: GPU RTX 2070 super My result of validation is almost 0 (mAP 0.001). There is no error during training. Is there any suggestion? Thank you.

Czm369 commented 3 months ago

More details, such as config and log, please.

gaby20000413 commented 3 months ago

Config is the same as yours, and no changes have been made. This is my train log: 20240312_204710.log Thank you for your reply.

PAGF188 commented 3 months ago

The Faster RCNN config (config) has an error and perform the evaluation each 50 training iterations:

https://github.com/Czm369/MixPL/blob/9f2176a3e4df8bcc12e8dc0be2e62decbaae9781/projects/MixPL/configs/mixpl_faster-rcnn_r50-caffe_fpn_180k_coco-s1-p10.py#L70-L72

The bbox_mAP you are getting is the AP after training the model in only 1100 iterations out of the 180K needed (that's why you get 0).

Czm369 commented 3 months ago

val_interval=5000

tamama9018 commented 3 months ago

Thanks for your great work! I tried the exact same method in 180K interations (I changed only val_interval=5000) and mAP gave the following results.

2024/03/15 20:47:31 - mmengine - INFO - bbox_mAP_copypaste: 0.112 0.226 0.097 0.054 0.123 0.151
2024/03/15 20:47:31 - mmengine - INFO - Iter(val) [5000/5000]    teacher/coco/bbox_mAP: 0.1490  teacher/coco/bbox_mAP_50: 0.2720  teacher/coco/bbox_mAP_75: 0.1480  teacher/coco/bbox_mAP_s: 0.0790  teacher/coco/bbox_mAP_m: 0.1580  teacher/coco/bbox_mAP_l: 0.2010  student/coco/bbox_mAP: 0.1120  student/coco/bbox_mAP_50: 0.2260  student/coco/bbox_mAP_75: 0.0970  student/coco/bbox_mAP_s: 0.0540  student/coco/bbox_mAP_m: 0.1230  student/coco/bbox_mAP_l: 0.1510  data_time: 0.0071  time: 0.0399
2024/03/15 20:47:31 - mmengine - INFO - Saving checkpoint at 1 epochs

This is my train log: 20240314_032925.log

I don't seem to have reached the mAP described in the paper (37.16 ± 0.15 ), am I doing it right? I would be happy to receive a reply.