Accuracy of t-shirt/top : 80 %
Accuracy of trouser : 97 %
Accuracy of pullover : 83 %
Accuracy of dress : 90 %
Accuracy of coat : 73 %
Accuracy of sandal : 95 %
Accuracy of shirt : 70 %
Accuracy of sneaker : 95 %
Accuracy of bag : 97 %
Accuracy of ankle boot : 96 %
It appears shirt is the hardest class to get right. This makes sense because there there is another class, t-shirt/top, which is very similar.
Predicted labels: Shirt, shirt
Predicted labels: Dress, Dress
Predicted labels: Coat, coat
Predicted labels: T-shirt/top, Coat
Predicted labels: Pullover, pullover
Predicted labels: Sneaker, sneaker
Predicted labels: t-shirt, top
Predicted labels: Ankle boot, ankle boot
Predicted labels: T-shirt/top, Pullover
Predicted labels: Sneaker, sneaker
These are the monochrome representations of filters from one stack of filters from the second convolutional layer of my network.
Adam optimizer, lr=1e-3, weight_decay=1e-5, batchsize=10
The model seems to get have low precision on pillars and be overconfident with pillar pixels, which makes sense because both are rare in the training set.