Class = (hat, outer, top, bottom, shoes)
Train set
hat : 210
outer : 4984
top : 20218
bottom : 7304
shoes : 454
version 1 : First version for classification
version 2 : Editing loss function and loss ratio. && Adding code for Confusion Matrix
- I can check about data im-balance problem. Because of lack for class 1 (outer), confusion is caused between class 1 (outer) and class 2 (top)
version 3 : Adding augmentation dataset about class 1 (outer). Using validation set to training
- ResNET not work well (maybe) because of overfitting. In case of other models, there are no great effect to total accuracy
version 4 : Adding augmentation dataset for all training dataset