about DISFA F1score

[vhzy]

Hello! I followed the image example you provided for pre-processing and got the same image database.
However after completing the training for stage2 using resnet50, the three folds average is less than 60.
I used your model for test and found it to be consistent with what the paper reported, even higher actually.
Here are some details:
Using the second fold (provided in your code) for training, the highest score was probably about 56 after completing stage 1 and about 58 after completing stage 2.
I wonder if i missed something in the training process?Thank you!

[yangchengjun]

I also have very low accuracy, even with the author's model, and the command and F1 scores are as follows:

python test.py --dataset DISFA --arc resnet50 --exp-name MEflod1_ResNet50_in_DISFA --fold 2  --resume results/ME-GraphAU_resnet50_DISFA/MEFARG_resnet50_DISFA_fold1.pth

2024-01-30 23:50:26,927:INFO: {'val_mean_f1_score 40.76 val_mean_acc 83.07'}
2024-01-30 23:50:26,927:INFO: {'F1-score-list:'}
2024-01-30 23:50:26,927:INFO: {'AU1: 20.54 AU2: 2.18 AU4: 64.23 AU6: 69.65 AU9: 45.61 AU12: 39.35 AU25: 52.82 AU26: 31.67 '}
2024-01-30 23:50:26,927:INFO: {'Acc-list:'}
2024-01-30 23:50:26,927:INFO: {'AU1: 77.06 AU2: 96.91 AU4: 84.93 AU6: 95.05 AU9: 96.36 AU12: 89.27 AU25: 46.12 AU26: 78.88 '}

python test.py --dataset DISFA --arc resnet50 --exp-name Test_MEflod2_ResNet50_in_DISFA --fold 2  --resume results/ME-GraphAU_resnet50_DISFA/MEFARG_resnet50_DISFA_fold2.pth

{'val_mean_f1_score 19.68 val_mean_acc 90.37'}
{'F1-score-list:'}
{'AU1: 13.39 AU2: 0.00 AU4: 11.35 AU6: 22.51 AU9: 0.00 AU12: 38.10 AU25: 70.18 AU26: 1.92 '}
{'Acc-list:'}
{'AU1: 93.32 AU2: 97.98 AU4: 84.49 AU6: 91.23 AU9: 97.05 AU12: 83.29 AU25: 84.94 AU26: 90.63 '}

python test.py --dataset DISFA --arc resnet50 --exp-name Test_MEflod3_ResNet50_in_DISFA --fold 2  --resume results/ME-GraphAU_resnet50_DISFA/MEFARG_resnet50_DISFA_fold3.pth

{'val_mean_f1_score 39.95 val_mean_acc 86.33'}
{'F1-score-list:'}
{'AU1: 21.50 AU2: 26.07 AU4: 43.63 AU6: 53.44 AU9: 39.78 AU12: 43.95 AU25: 69.83 AU26: 21.38 '}
{'Acc-list:'}
{'AU1: 77.82 AU2: 96.38 AU4: 81.47 AU6: 92.63 AU9: 97.46 AU12: 86.98 AU25: 76.13 AU26: 81.78 '}