nb_classes = 9


confusion_matrix = torch.zeros(nb_classes, nb_classes)
cat1_corr = 0
cat1 = 0
cat2_corr = 0
cat2 = 0
cat3_corr = 0
cat3 = 0
cat4_corr = 0
cat4 = 0
cat5_corr = 0
cat5 = 0
cat6_corr = 0
cat6 = 0
cat7_corr = 0
cat7 = 0
cat8_corr = 0
cat8 = 0
cat9_corr = 0
cat9 = 0
_classes = []
predicted_labels = []
with torch.no_grad():
    for i, (inputs, classes) in enumerate(dataloaders['test']):


        inputs = inputs.to(device)
        tmp_labels = model_ft(inputs)

        classes = classes.to(device)
        classes_list = classes.cpu().detach().numpy().tolist()
        print('classes list: ', classes_list)
        _classes[:]=[i+1 for i in classes_list]
        print('_classes: ', _classes)
        preds = preds.cpu().detach().numpy().tolist()
        outputs = model_ft(inputs)
        _, preds = torch.max(outputs, 1)
        preds_list = preds.cpu().detach().numpy().tolist()
        _preds[:]=[i+1 for i in preds_list]
        for i in range(4):
            if _classes[i] == 1:
                cat1 += 1
                if _classes[i] == _preds[i]:
                    cat1_corr += 1
            elif _classes[i] == 2:
                cat2 += 1
                if _classes[i] == _preds[i]:
                    cat2_corr += 1
            elif _classes[i] == 3:
                cat3 += 1
                if _classes[i] == _preds[i]:
                    cat3_corr += 1
            elif _classes[i] == 4:
                cat4 += 1
                if _classes[i] == _preds[i]:
                    cat4_corr += 1
            elif _classes[i] == 5:
                cat5 += 1
                if _classes[i] == _preds[i]:
                    cat5_corr += 1
            elif _classes[i] == 6:
                cat6 += 1
                if _classes[i] == _preds[i]:
                    cat6_corr += 1
            elif _classes[i] == 7:
                cat7 += 1
                if _classes[i] == _preds[i]:
                    cat7_corr += 1
            elif _classes[i] == 8:
                cat8 += 1
                if _classes[i] == _preds[i]:
                    cat8_corr += 1
            elif _classes[i] == 9:
                print('here')
                cat9 += 1
                if _classes[i] == _preds[i]:
                    cat9_corr += 1


        predicted_labels.append(preds.cpu().detach().numpy().tolist())
        for t, p in zip(classes.view(-1), preds.view(-1)):
                confusion_matrix[t.long(), p.long()] += 1

print(confusion_matrix)


cat1_acc = cat1_corr/cat1
cat2_acc = cat2_corr/cat2
cat3_acc = cat3_corr/cat3
cat4_acc = cat4_corr/cat4
cat5_acc = cat5_corr/cat5
cat6_acc = cat6_corr/cat6
cat7_acc = cat7_corr/cat7
cat8_acc = cat8_corr/cat8
cat9_acc = cat9_corr/cat9


print('1', cat1_acc)
print('2', cat2_acc)
print('3', cat3_acc)
print('4', cat4_acc)
print('5', cat5_acc)
print('6', cat6_acc)
print('7', cat7_acc)
print('8', cat8_acc)
print('9', cat9_acc)