Kaggle Competition: https://www.kaggle.com/c/plant-pathology-2020-fgvc7
framework: pytorch
Data Augmentation uses RandomHorizontalFlip, RandomSizedCrop, RandomRotation.
I used some pre-trained models, such as efficientnet-b0, efficientnet-b3, efficientnet-b6, inception_v3, resnet101.
In addition, I set 5 k-fold to help increase accuracy.
-Learning rate setting:
-test time augmentation (TTA): In order to improve the accuracy of the prediction results, I use 8 TTA.
-Parameters:
Image size: 256, 512
Batch size: 16, 32
-Confusion Matrix
-Training & Validation Accuracy
-Training & Validation Loss
