aug data on tensorboard

@@ -33,11 +33,12 @@ def eval(model_path):
     print('\n[+] Load dataset')
     print('\n[+] Load dataset')
     test_transform = get_valid_transform(args, model)
     test_transform = get_valid_transform(args, model)
     test_dataset = get_dataset(args, test_transform, 'test')
     test_dataset = get_dataset(args, test_transform, 'test')
-    test_loader = iter(get_dataloader(args, test_dataset))
+    print("len(dataset): ",  len(test_dataset), type(test_dataset)) # 590
+    
+    test_loader = iter(get_dataloader(args, test_dataset)) ###
 
 
     print('\n[+] Start testing')
     print('\n[+] Start testing')
-    log_dir = os.path.join('/content/drive/My Drive/CD2 Project/runs', model_name)
-    writer = SummaryWriter(log_dir=log_dir)
+    writer = SummaryWriter(log_dir=model_path)
     _test_res = validate(args, model, criterion, test_loader, step=0, writer=writer)
     _test_res = validate(args, model, criterion, test_loader, step=0, writer=writer)
 
 
     print('\n[+] Valid results')
     print('\n[+] Valid results')

@@ -55,7 +55,7 @@ def split_dataset(args, dataset, k):
 
 
     return Dm_indexes, Da_indexes
     return Dm_indexes, Da_indexes
 
 
-
+#(images[j], first[j]), global_step=step)
 def concat_image_features(image, features, max_features=3):
 def concat_image_features(image, features, max_features=3):
     _, h, w = image.shape
     _, h, w = image.shape
 
 
@@ -93,6 +93,7 @@ def concat_image_features(image, features, max_features=3):
         
         
 
 
         image_feature = torch.cat((image_feature, feature), 2) ### dim = 2
         image_feature = torch.cat((image_feature, feature), 2) ### dim = 2
+        #print("\nimg feature size: ", image_feature.size()) #[1, 240, 720]
 
 
     return image_feature
     return image_feature
 
 
@@ -149,10 +150,6 @@ def parse_args(kwargs):
 
 
 
 
 def select_model(args):
 def select_model(args):
-    # resnet_dict = {'ResNet18':grayResNet.ResNet18(), 'ResNet34':grayResNet.ResNet34(), 
-    # 'ResNet50':grayResNet.ResNet50(), 'ResNet101':grayResNet.ResNet101(), 'ResNet152':grayResNet.ResNet152()}
-    
-
     # grayResNet2
     # grayResNet2
     resnet_dict = {'resnet18':grayResNet2.resnet18(), 'resnet34':grayResNet2.resnet34(), 
     resnet_dict = {'resnet18':grayResNet2.resnet18(), 'resnet34':grayResNet2.resnet34(), 
     'resnet50':grayResNet2.resnet50(), 'resnet101':grayResNet2.resnet101(), 'resnet152':grayResNet2.resnet152()}
     'resnet50':grayResNet2.resnet50(), 'resnet101':grayResNet2.resnet101(), 'resnet152':grayResNet2.resnet152()}
@@ -285,7 +282,7 @@ def get_dataset(args, transform, split='train'):
     elif args.dataset == 'BraTS':
     elif args.dataset == 'BraTS':
       if split in ['train']:
       if split in ['train']:
         dataset = CustomDataset(TRAIN_DATASET_PATH, transform=transform)
         dataset = CustomDataset(TRAIN_DATASET_PATH, transform=transform)
-      else:
+      else: #test
         dataset = CustomDataset(VAL_DATASET_PATH, transform=transform)
         dataset = CustomDataset(VAL_DATASET_PATH, transform=transform)
       
       
 
 
@@ -382,7 +379,7 @@ def get_valid_transform(args, model):
             transforms.ToTensor()
             transforms.ToTensor()
         ])
         ])
     elif args.dataset == 'BraTS':
     elif args.dataset == 'BraTS':
-        resize_h, resize_w = 256, 256
+        resize_h, resize_w = 240, 240
         val_transform = transforms.Compose([
         val_transform = transforms.Compose([
             transforms.Resize([resize_h, resize_w]),
             transforms.Resize([resize_h, resize_w]),
             transforms.ToTensor()
             transforms.ToTensor()
@@ -426,13 +423,14 @@ def train_step(args, model, optimizer, scheduler, criterion, batch, step, writer
 
 
     if writer and step % args.print_step == 0:
     if writer and step % args.print_step == 0:
         n_imgs = min(images.size(0), 10)
         n_imgs = min(images.size(0), 10)
+        tag = 'train/' + str(step)
         for j in range(n_imgs):
         for j in range(n_imgs):
-            writer.add_image('train/input_image',
+            writer.add_image(tag,
                     concat_image_features(images[j], first[j]), global_step=step)
                     concat_image_features(images[j], first[j]), global_step=step)
 
 
     return acc1, acc5, loss, forward_t, backward_t
     return acc1, acc5, loss, forward_t, backward_t
 
 
-
+# validate(args, model, criterion, test_loader, step=0, writer=writer)
 def validate(args, model, criterion, valid_loader, step, writer, device=None):
 def validate(args, model, criterion, valid_loader, step, writer, device=None):
     # switch to evaluate mode
     # switch to evaluate mode
     model.eval()
     model.eval()
@@ -441,19 +439,24 @@ def validate(args, model, criterion, valid_loader, step, writer, device=None):
     samples = 0
     samples = 0
     infer_t = 0
     infer_t = 0
 
 
+    img_count = 0
+
     with torch.no_grad():
     with torch.no_grad():
-        for i, (images, target) in enumerate(valid_loader): ##
+        for i, (images, target) in enumerate(valid_loader): ## loop [0, 148]
 
 
+            #print("\n1 images size: ", images.size()) #[4, 1, 240, 240]
             start_t = time.time()
             start_t = time.time()
             if device:
             if device:
                 images = images.to(device)
                 images = images.to(device)
                 target = target.to(device)
                 target = target.to(device)
 
 
-            elif args.use_cuda is not None:
+            elif args.use_cuda is not None: #
                 images = images.cuda(non_blocking=True)
                 images = images.cuda(non_blocking=True)
                 target = target.cuda(non_blocking=True)
                 target = target.cuda(non_blocking=True)
+            #print("\n2 images size: ", images.size()) #[4, 1, 240, 240]
 
 
             # compute output
             # compute output
+            # first = nn.Sequential(*list(backbone.children())[:1])
             output, first = model(images)
             output, first = model(images)
             loss = criterion(output, target)
             loss = criterion(output, target)
             infer_t += time.time() - start_t
             infer_t += time.time() - start_t
@@ -464,14 +467,19 @@ def validate(args, model, criterion, valid_loader, step, writer, device=None):
             acc5 += _acc5
             acc5 += _acc5
             samples += images.size(0)
             samples += images.size(0)
 
 
+            if writer:
+               # print("\n3 images.size(0): ", images.size(0))
+                n_imgs = min(images.size(0), 10)
+                for j in range(n_imgs):
+                    tag = 'valid/' + str(img_count)
+                    writer.add_image(tag,
+                            concat_image_features(images[j], first[j]), global_step=step)    
+                    img_count = img_count + 1
+
     acc1 /= samples
     acc1 /= samples
     acc5 /= samples
     acc5 /= samples
 
 
-    if writer:
-        n_imgs = min(images.size(0), 10)
-        for j in range(n_imgs):
-            writer.add_image('valid/input_image',
-                    concat_image_features(images[j], first[j]), global_step=step)
+  
 
 
     return acc1, acc5, loss, infer_t
     return acc1, acc5, loss, infer_t