futures.md

1.

def plot_filtered_images(images, filters):

    images = torch.cat([i.unsqueeze(0) for i in images], dim=0).cpu()
    
    print(images.shape)
    
    filters = filters.cpu()

    n_images  = images.shape[0]
    n_filters = filters.shape[0]

    filtered_images = F.conv2d(images, filters)
    
    fig = plt.figure(figsize=(20, 20))

    for i in range(n_images):

        img = images[i].squeeze(0).numpy().transpose((1,2,0))
        
        img = 255*(img - img.min())/(img.max() - img.min())

        ax = fig.add_subplot(n_images, n_filters+1, i+1+(i*n_filters))
        
        ax.imshow(img.astype(int), cmap='bone')
        ax.set_title('Original')
        ax.axis('off')

        for j in range(n_filters):
            image = filtered_images[i][j]
            
            image = image.numpy().astype(float)
            
            image = 255*(image - image.min())/(image.max() - image.min())
            
            ax = fig.add_subplot(n_images, n_filters+1, i+1+(i*n_filters)+j+1)
            
            ax.imshow(image.astype(int), cmap='bone')
            
            ax.set_title(f'Filter {j+1}')
            ax.axis('off')
    return filtered_images

N_IMAGES = 5

images = [image for image, label in [testset[i] for i in range(N_IMAGES)]]

filters = model.conv1.weight.data

filtered_images = plot_filtered_images(images, filters)

2.

n_classes = 10
n_samples = len(testset)

class_correct = torch.zeros(n_classes)
class_total   = torch.zeros(n_classes)

labels   = []
predicts = []

model.eval()

with torch.no_grad():
    
    for x,y in tqdm(testloader):
        x = x.to(device)
        y = y.to('cpu')

        y_pred = model.forward(x).to('cpu')
        
        cls_pred = torch.argmax(y_pred, dim=1)
        
        c = (cls_pred == y).squeeze() # one-hot vector
        
        for i in range(x.shape[0]):
            label = y[i]
            class_correct[label] += c[i].item()
            class_total[label]   += 1
    
        labels.append(y)
        predicts.append(cls_pred)
        
labels   = torch.cat(labels, dim = 0) # make it 1d array
predicts = torch.cat(predicts, dim = 0) # make it 1d array
    
for i in range(n_classes):
    print('Accuracy of %5s : %2d %%' % (
        names_classes[i], 100 * class_correct[i] / class_total[i]))

sorted_ids = list(range(1, n_samples + 1)) 

sorted_ids.sort(key=lambda x: str(x))

df = pd.DataFrame({'id': sorted_ids, 'label': predicts})

df['label'] = df['label'].apply(lambda x: names_classes[x])

df.to_csv('submission.csv', index=False)