search

caoyunkang / CDO

[TII 2023] Collaborative Discrepancy Optimization for Reliable Image Anomaly Localization
MIT License
61 stars 7 forks source link

About AU-Pro metrics #9

Open Jay-zzcoder opened 7 months ago

Jay-zzcoder commented 7 months ago

Thank you for your excellent work! But I meet some problems when I run this code: Why the metrics AU-Pro is always 0.0 when training like this: 2024-01-30 12:55:23.070 | INFO | main:main:135 - bagel=======i_roc: 96.44 2024-01-30 12:55:23.071 | INFO | main:main:135 - bagel=======p_roc: 99.23 2024-01-30 12:55:23.071 | INFO | main:main:135 - bagel=======p_pro: 0.00 2024-01-30 12:55:23.072 | INFO | main:main:135 - bagel=======threshold: 0.75

caoyunkang commented 7 months ago

hi, the calculation process of AU-PRO is really time-consuming, so I discarded AU-PRO in the training process. Please carefully check the flag cal_pro for further information.

Jay-zzcoder commented 7 months ago

hi, the calculation process of AU-PRO is really time-consuming, so I discarded AU-PRO in the training process. Please carefully check the flag cal_pro for further information.

Whether the flag 'cal_pro' is set as True or Fasle, AU_Pro is always0.0 2024-01-31 01:21:15.018 | INFO | main:main:74 - ==========running parameters============= 2024-01-31 01:21:15.019 | INFO | main:main:76 - dataset: mvtec3d 2024-01-31 01:21:15.019 | INFO | main:main:76 - class_name: bagel 2024-01-31 01:21:15.019 | INFO | main:main:76 - img_resize: 256 2024-01-31 01:21:15.019 | INFO | main:main:76 - img_cropsize: 256 2024-01-31 01:21:15.019 | INFO | main:main:76 - num_epochs: 2 2024-01-31 01:21:15.019 | INFO | main:main:76 - validation_epoch: 5 2024-01-31 01:21:15.019 | INFO | main:main:76 - lr: 0.0004 2024-01-31 01:21:15.019 | INFO | main:main:76 - batch_size: 16 2024-01-31 01:21:15.019 | INFO | main:main:76 - vis: True 2024-01-31 01:21:15.019 | INFO | main:main:76 - root_dir: ./result 2024-01-31 01:21:15.019 | INFO | main:main:76 - load_memory: True 2024-01-31 01:21:15.019 | INFO | main:main:76 - cal_pro: True 2024-01-31 01:21:15.019 | INFO | main:main:76 - seed: 111 2024-01-31 01:21:15.019 | INFO | main:main:76 - gpu_id: 0 2024-01-31 01:21:15.019 | INFO | main:main:76 - pure_test: False 2024-01-31 01:21:15.019 | INFO | main:main:76 - backbone: hrnet32 2024-01-31 01:21:15.020 | INFO | main:main:76 - MOM: True 2024-01-31 01:21:15.020 | INFO | main:main:76 - OOM: True 2024-01-31 01:21:15.020 | INFO | main:main:76 - gamma: 2.0 2024-01-31 01:21:15.020 | INFO | main:main:77 - ========================================= 2024-01-31 01:21:15.026 | INFO | utils.training_utils:get_dir_from_args:69 - ===> Root dir for this experiment: ./result/hrnet32-mvtec3d-wi-OOM-wi-MOM/logger/bagel 2024-01-31 01:21:17.217 | INFO | datasets:get_dataloader_from_args:33 - ===> datasets: mvtec3d, class name/len: bagel/110, batch size: 16 2024-01-31 01:21:23.810 | INFO | datasets:get_dataloader_from_args:33 - ===> datasets: mvtec3d, class name/len: bagel/244, batch size: 16 CDO:bagel: 0%| | 0/2 [00:00<?, ?it/s]2024-01-31 01:21:37.665 | INFO | main:main:131 -

2024-01-31 01:21:37.667 | INFO | main:main:135 - bagel=======i_roc: 57.28 2024-01-31 01:21:37.668 | INFO | main:main:135 - bagel=======p_roc: 94.22 2024-01-31 01:21:37.668 | INFO | main:main:135 - bagel=======p_pro: 0.00 2024-01-31 01:21:37.669 | INFO | main:main:135 - bagel=======threshold: 1.59