added result, 21.9691

image-inpainting/src/architecture.py

@@ -15,8 +15,10 @@ def init_weights(m):
         nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
         if m.bias is not None:
             nn.init.constant_(m.bias, 0)
-    elif isinstance(m, nn.BatchNorm2d):
+    elif isinstance(m, (nn.BatchNorm2d, nn.InstanceNorm2d)):
+        if m.weight is not None:
             nn.init.constant_(m.weight, 1)
+        if m.bias is not None:
             nn.init.constant_(m.bias, 0)
 
 
@@ -69,35 +71,36 @@ class CBAM(nn.Module):
 
 
 class ConvBlock(nn.Module):
-    """Convolutional block with Conv2d -> BatchNorm -> LeakyReLU"""
+    """Convolutional block with Conv2d -> InstanceNorm2d -> GELU"""
-    def __init__(self, in_channels, out_channels, kernel_size=3, padding=1, dropout=0.0):
+    def __init__(self, in_channels, out_channels, kernel_size=3, padding=1, dropout=0.0, dilation=1):
         super().__init__()
-        self.conv = nn.Conv2d(in_channels, out_channels, kernel_size, padding=padding)
+        self.conv = nn.Conv2d(in_channels, out_channels, kernel_size, padding=padding, dilation=dilation)
-        self.bn = nn.BatchNorm2d(out_channels)
+        # InstanceNorm is preferred for style/inpainting tasks
-        self.relu = nn.LeakyReLU(0.1, inplace=True)
+        self.bn = nn.InstanceNorm2d(out_channels, affine=True)
+        self.act = nn.GELU()
         self.dropout = nn.Dropout2d(dropout) if dropout > 0 else nn.Identity()
     
     def forward(self, x):
-        return self.dropout(self.relu(self.bn(self.conv(x))))
+        return self.dropout(self.act(self.bn(self.conv(x))))
 
 class ResidualConvBlock(nn.Module):
     """Residual convolutional block for better gradient flow"""
-    def __init__(self, channels, dropout=0.0):
+    def __init__(self, channels, dropout=0.0, dilation=1):
         super().__init__()
-        self.conv1 = nn.Conv2d(channels, channels, 3, padding=1)
+        self.conv1 = nn.Conv2d(channels, channels, 3, padding=dilation, dilation=dilation)
-        self.bn1 = nn.BatchNorm2d(channels)
+        self.bn1 = nn.InstanceNorm2d(channels, affine=True)
-        self.conv2 = nn.Conv2d(channels, channels, 3, padding=1)
+        self.conv2 = nn.Conv2d(channels, channels, 3, padding=dilation, dilation=dilation)
-        self.bn2 = nn.BatchNorm2d(channels)
+        self.bn2 = nn.InstanceNorm2d(channels, affine=True)
-        self.relu = nn.LeakyReLU(0.1, inplace=True)
+        self.act = nn.GELU()
         self.dropout = nn.Dropout2d(dropout) if dropout > 0 else nn.Identity()
     
     def forward(self, x):
         residual = x
-        out = self.relu(self.bn1(self.conv1(x)))
+        out = self.act(self.bn1(self.conv1(x)))
         out = self.dropout(out)
         out = self.bn2(self.conv2(out))
         out = out + residual
-        return self.relu(out)
+        return self.act(out)
 
 
 class DownBlock(nn.Module):
@@ -161,9 +164,9 @@ class MyModel(nn.Module):
         # Bottleneck with multiple residual blocks
         self.bottleneck = nn.Sequential(
             ConvBlock(base_channels * 16, base_channels * 16, dropout=dropout),
-            ResidualConvBlock(base_channels * 16, dropout=dropout),
+            ResidualConvBlock(base_channels * 16, dropout=dropout, dilation=2),
-            ResidualConvBlock(base_channels * 16, dropout=dropout),
+            ResidualConvBlock(base_channels * 16, dropout=dropout, dilation=4),
-            ResidualConvBlock(base_channels * 16, dropout=dropout),
+            ResidualConvBlock(base_channels * 16, dropout=dropout, dilation=8),
             CBAM(base_channels * 16)
         )
         
@@ -183,7 +186,7 @@ class MyModel(nn.Module):
         # Output layer with smooth transition
         self.output = nn.Sequential(
             nn.Conv2d(base_channels, base_channels // 2, kernel_size=3, padding=1),
-            nn.LeakyReLU(0.1, inplace=True),
+            nn.GELU(),
             nn.Conv2d(base_channels // 2, 3, kernel_size=1),
             nn.Sigmoid()  # Ensure output is in [0, 1] range
         )

image-inpainting/src/datasets.py

@@ -26,11 +26,21 @@ def create_arrays_from_image(image_array: np.ndarray, offset: tuple, spacing: tu
 
     return image_array, known_array
 
-def resize(img: Image):
+def resize(img: Image, augment: bool = False):
-    resize_transforms = transforms.Compose([
+    transforms_list = [
         transforms.Resize((IMAGE_DIMENSION, IMAGE_DIMENSION)),
         transforms.CenterCrop((IMAGE_DIMENSION, IMAGE_DIMENSION))
-    ])
+    ]
+    
+    if augment:
+        transforms_list = [
+            transforms.RandomHorizontalFlip(),
+            transforms.RandomVerticalFlip(),
+            transforms.ColorJitter(brightness=0.1, contrast=0.1, saturation=0.1, hue=0.05),
+            transforms.RandomRotation(10),
+        ] + transforms_list
+        
+    resize_transforms = transforms.Compose(transforms_list)
     return resize_transforms(img)
 def preprocess(input_array: np.ndarray):
     input_array = np.asarray(input_array, dtype=np.float32) / 255.0
@@ -41,8 +51,9 @@ class ImageDataset(torch.utils.data.Dataset):
     Dataset class for loading images from a folder
     """
 
-    def __init__(self, datafolder: str):
+    def __init__(self, datafolder: str, augment: bool = False):
         self.imagefiles = sorted(glob.glob(os.path.join(datafolder,"**","*.jpg"),recursive=True))
+        self.augment = augment
 
     def __len__(self):
         return len(self.imagefiles)
@@ -51,7 +62,7 @@ class ImageDataset(torch.utils.data.Dataset):
         index = int(idx)
         
         image = Image.open(self.imagefiles[index])
-        image = np.asarray(resize(image))
+        image = np.asarray(resize(image, self.augment))
         image = preprocess(image)
         spacing_x = random.randint(2,6)
         spacing_y = random.randint(2,6)

image-inpainting/src/train.py

@@ -84,16 +84,21 @@ def train(seed, testset_ratio, validset_ratio, data_path, results_path, early_st
     plotpath = os.path.join(results_path, "plots")
     os.makedirs(plotpath, exist_ok=True)
 
-    image_dataset = datasets.ImageDataset(datafolder=data_path)
+    image_dataset = datasets.ImageDataset(datafolder=data_path, augment=False)
 
     n_total = len(image_dataset)
     n_test = int(n_total * testset_ratio)
     n_valid = int(n_total * validset_ratio)
     n_train = n_total - n_test - n_valid
     indices = np.random.permutation(n_total)
-    dataset_train = Subset(image_dataset, indices=indices[0:n_train])
+    
-    dataset_valid = Subset(image_dataset, indices=indices[n_train:n_train + n_valid])
+    # Create datasets with and without augmentation
-    dataset_test = Subset(image_dataset, indices=indices[n_train + n_valid:n_total])
+    train_dataset_source = datasets.ImageDataset(datafolder=data_path, augment=True)
+    val_test_dataset_source = datasets.ImageDataset(datafolder=data_path, augment=False)
+    
+    dataset_train = Subset(train_dataset_source, indices=indices[0:n_train])
+    dataset_valid = Subset(val_test_dataset_source, indices=indices[n_train:n_train + n_valid])
+    dataset_test = Subset(val_test_dataset_source, indices=indices[n_train + n_valid:n_total])
 
     assert len(image_dataset) == len(dataset_train) + len(dataset_test) + len(dataset_valid)