classificação qualidade

tl;dr:
- executado YOLO em cima da "amostra boa" e "amostra ruim"
- labels = Bad/Good
- abordagem simples, VGG16 5-fold cross-validation
- accuracy 78.13% (3097/3964)



detecções YOLO

Puruna_1-Bad
Detected: 433/1983 (21.84)
Narinas: [1035,514,433,1,0]
Tempo: 905.15

Puruna_2-Bad
Detected: 678/2900 (23.38)
Narinas: [1242,980,678,0,0]
Tempo: 1242.00

Puruna_3-Bad
Detected: 977/3773 (25.89)
Narinas: [1674,1122,977,0,0]
Tempo: 1763.52


Puruna_1-Good
Detected: 442/722 (61.22)
Narinas: [109,171,442,0,0]
Tempo: 356.50

Puruna_2-Good
Detected: 618/1126 (54.88)
Narinas: [141,365,618,2,0]
Tempo: 531.47

Puruna_3-Good
Detected: 816/1403 (58.16)
Narinas: [153,434,816,0,0]
Tempo: 634.65



5-fold CV

Train feature extraction finished in 273.651516 seconds.
Test feature extraction finished in 68.686547 seconds.
              precision    recall  f1-score   support

         Bad       0.82      0.63      0.71       418
        Good       0.67      0.85      0.75       376

    accuracy                           0.73       794
   macro avg       0.75      0.74      0.73       794
weighted avg       0.75      0.73      0.73       794

Acc: 0.7342569269521411 794 583
Train Feature Extraction...
Train feature extraction finished in 274.474096 seconds.
Test feature extraction finished in 68.600150 seconds.
              precision    recall  f1-score   support

         Bad       0.80      0.73      0.76       418
        Good       0.73      0.80      0.76       375

    accuracy                           0.76       793
   macro avg       0.76      0.76      0.76       793
weighted avg       0.77      0.76      0.76       793

Acc: 0.7629255989911727 793 605
Train Feature Extraction...
Train feature extraction finished in 274.467754 seconds.
Test feature extraction finished in 68.850336 seconds.
              precision    recall  f1-score   support

         Bad       0.85      0.78      0.81       418
        Good       0.77      0.84      0.81       375

    accuracy                           0.81       793
   macro avg       0.81      0.81      0.81       793
weighted avg       0.81      0.81      0.81       793

Acc: 0.8083228247162674 793 641
Train Feature Extraction...
Train feature extraction finished in 274.260627 seconds.
Test feature extraction finished in 68.514654 seconds.
              precision    recall  f1-score   support

         Bad       0.80      0.76      0.78       417
        Good       0.75      0.79      0.77       375

    accuracy                           0.77       792
   macro avg       0.77      0.77      0.77       792
weighted avg       0.78      0.77      0.77       792

Acc: 0.773989898989899 792 613
Train Feature Extraction...
Train feature extraction finished in 273.932529 seconds.
Test feature extraction finished in 68.439771 seconds.
              precision    recall  f1-score   support

         Bad       0.85      0.81      0.83       417
        Good       0.80      0.84      0.82       375

    accuracy                           0.83       792
   macro avg       0.83      0.83      0.83       792
weighted avg       0.83      0.83      0.83       792

Acc: 0.827020202020202 792 655
End: 3097.0 3964 0.7812815338042381