Add files via upload

exploring_data/dataloader_demo.py

@@ -0,0 +1,25 @@
+from out import get_data_loaders
+import torch 
+
+def loaders_demo():
+    full_dataloaders, _ = get_data_loaders(
+    {'train': './data/1000_train_mnistmnistmsvhnsynusps.npz', 
+     'test': './data/1000_test_mnistmnistmsvhnsynusps.npz',
+     },
+    batch_size= 64) 
+    print(full_dataloaders.keys())
+
+    for phase in ['train', 'test', 'test_missing']: 
+        print(f'{phase} data...')
+        for batch_indx, (images, features, domain_labels, digit_labels) in enumerate(full_dataloaders[phase]):
+            print(f'{batch_indx}-th batch')
+            print('images shape: ', images.shape)
+            print('features shape: ', features.shape)
+            if phase == 'test_missing': 
+                print('in test-missing dataloaders, since the features are not available, features are filled with zeros', torch.sum(features))
+            print('domain labels freq: ', torch.unique(domain_labels, return_counts=True))
+            print('digit labels freq: ', torch.unique(digit_labels, return_counts=True))
+            print()
+            break
+
+loaders_demo()

exploring_data/exploring_data.py

@@ -0,0 +1,94 @@
+import torch
+import numpy as np
+import matplotlib.pyplot as plt
+from out import get_data_loaders
+
+NUM_SAMPLES_FROM_DIGITS = 5
+NUM_DOMAINS = 5
+NUM_DIGITS = 10
+SEED = 141+1
+
+DOMAIN_NAME_DICT = {
+    0: 'mnist',
+    1: 'mnistm',
+    2: 'svhn',
+    3: 'syn',
+    4: 'usps' 
+} 
+
+# get data loaders
+# 1000_test_mnistmnistmsvhnsynusps.npz
+# 1000_train_mnistmnistmsvhnsynusps.npz
+full_dataloaders, _ = get_data_loaders(
+    {'train': './data/1000_train_mnistmnistmsvhnsynusps.npz', 
+     'test': './data/1000_test_mnistmnistmsvhnsynusps.npz',
+     },
+    batch_size= 64, init_seed=SEED) 
+
+# utils
+def get_samples(dataloader, num_samples: int) -> np.ndarray: 
+    samples = [
+        [ [], [], [], [], [], [], [], [], [], [] ],
+        [ [], [], [], [], [], [], [], [], [], [] ],
+        [ [], [], [], [], [], [], [], [], [], [] ],
+        [ [], [], [], [], [], [], [], [], [], [] ],
+        [ [], [], [], [], [], [], [], [], [], [] ],
+    ]
+
+    for batch_indx, (images, _, domain_labels, digit_labels) in enumerate(dataloader): 
+        # images are normalized using mean=(0.5, 0.5, 0.5) and std=(0.5, 0.5, 0.5),
+        # so images habve been normalized using: image = image - mean / std
+        # to plot images we have to undo the normalization 
+        images = images * 0.5 + 0.5
+
+        for img_indx, curr_image in enumerate(images):
+
+            if len(samples[domain_labels[img_indx]][digit_labels[img_indx]]) < num_samples:
+                samples[domain_labels[img_indx]][digit_labels[img_indx]].append(curr_image.numpy())
+
+    # convert samples to numpy array
+    return np.array(samples)
+
+
+
+# plot NUM_SAMPLES_FROM_DIGITS of each domain together: 
+samples = get_samples(full_dataloaders['train'], NUM_SAMPLES_FROM_DIGITS)
+
+fig_height = 10
+fig_width = fig_height * (NUM_DOMAINS*NUM_SAMPLES_FROM_DIGITS) / NUM_DIGITS
+fig, axs = plt.subplots(NUM_DIGITS, NUM_DOMAINS*NUM_SAMPLES_FROM_DIGITS, figsize=(fig_width, fig_height), 
+                        # gridspec_kw=dict(hspace=0.0)
+                        gridspec_kw={'height_ratios':[1]*10}
+                        )
+
+for i in range(NUM_DIGITS): 
+    for dom in range(NUM_DOMAINS):
+        for j in range(NUM_SAMPLES_FROM_DIGITS):  
+            image = np.transpose(samples[dom][i][j], (1, 2, 0))  # Transpose the image to (32, 32, 3)
+            axs[i, j+dom*NUM_SAMPLES_FROM_DIGITS].imshow(image)
+            axs[i, j+dom*NUM_SAMPLES_FROM_DIGITS].axis('off')
+
+plt.subplots_adjust(wspace=0, hspace=0, left=0, right=1, bottom=0, top=1)
+plt.tight_layout()
+plt.savefig('exploring_data/all_domains.jpg')
+plt.clf()
+
+
+# plot 10 samples of each domain separately
+samples = get_samples(full_dataloaders['train'], 10)
+
+for domain in DOMAIN_NAME_DICT.keys(): 
+    print(f'plotting samples from domain {DOMAIN_NAME_DICT[domain]}')
+
+    fig, axs = plt.subplots(10, 10, figsize=(10, 10), gridspec_kw=dict(hspace=0.0))
+
+    for i in range(10):  
+        for j in range(10):  
+            image = np.transpose(samples[domain][i][j], (1, 2, 0))  # Transpose the image to (32, 32, 3)
+            axs[j, i].imshow(image)
+            axs[j, i].axis('off')
+
+    plt.tight_layout()
+    plt.title(f'{DOMAIN_NAME_DICT[domain]}')
+    plt.savefig(f'exploring_data/{DOMAIN_NAME_DICT[domain]}.jpg')
+    plt.clf()