Update app.py

app.py

@@ -5,37 +5,42 @@ from PIL import Image
 import numpy as np
 import math
 import os
+from threading import Event
+import traceback
 
 # Constants
 IMG_SIZE = 128
-TIMESTEPS = 300
+TIMESTEPS = 300  # From second code
 NUM_CLASSES = 2
 
-# Device configuration
+# Global Cancellation Flag
+cancel_event = Event()
+
+# Device Configuration
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 
-# 1. Sinusoidal Embeddings
+# --- Model Definitions ---
 class SinusoidalPositionEmbeddings(nn.Module):
     def __init__(self, dim):
         super().__init__()
         self.dim = dim
         half_dim = dim // 2
         emb = math.log(10000) / (half_dim - 1)
-        emb = torch.exp(torch.arange(half_dim) * -emb)
+        emb = torch.exp(torch.arange(half_dim) * -emb)  # From second code (no dtype specified)
         self.register_buffer('embeddings', emb)
 
     def forward(self, time):
-        device = time.device
+        device = time.device  # From second code
         embeddings = self.embeddings.to(device)
-        embeddings = time[:, None] * embeddings[None, :]
+        embeddings = time[:, None] * embeddings[None, :]  # From second code
         return torch.cat([embeddings.sin(), embeddings.cos()], dim=-1)
 
-# 2. UNet Model (matches your original architecture exactly)
 class UNet(nn.Module):
     def __init__(self, in_channels=3, out_channels=3, num_classes=2, time_dim=256):
         super().__init__()
         self.num_classes = num_classes
         self.label_embedding = nn.Embedding(num_classes, time_dim)
+
         self.time_mlp = nn.Sequential(
             SinusoidalPositionEmbeddings(time_dim),
             nn.Linear(time_dim, time_dim),
@@ -43,16 +48,16 @@ class UNet(nn.Module):
             nn.Linear(time_dim, time_dim)
         )
 
-        # Encoder (matches your original channel sizes)
+        # Encoder
         self.inc = self.double_conv(in_channels, 64)
         self.down1 = self.down(64 + time_dim * 2, 128)
         self.down2 = self.down(128 + time_dim * 2, 256)
         self.down3 = self.down(256 + time_dim * 2, 512)
 
-        # Bottleneck (matches your original)
+        # Bottleneck
         self.bottleneck = self.double_conv(512 + time_dim * 2, 1024)
 
-        # Decoder (matches your original upsampling structure)
+        # Decoder
         self.up1 = nn.ConvTranspose2d(1024, 256, kernel_size=2, stride=2)
         self.upconv1 = self.double_conv(256 + 256 + time_dim * 2, 256)
 
@@ -79,7 +84,6 @@ class UNet(nn.Module):
         )
 
     def forward(self, x, labels, time):
-        # Matches your original forward pass exactly
         label_indices = torch.argmax(labels, dim=1)
         label_emb = self.label_embedding(label_indices)
         t_emb = self.time_mlp(time)
@@ -118,15 +122,14 @@ class UNet(nn.Module):
 
         return self.outc(x)
 
-# 3. Diffusion Model (matches your original implementation)
 class DiffusionModel(nn.Module):
-    def __init__(self, model, timesteps=500, time_dim=256):
+    def __init__(self, model, timesteps=TIMESTEPS, time_dim=256):
         super().__init__()
         self.model = model
         self.timesteps = timesteps
         self.time_dim = time_dim
 
-        # Linear beta schedule (matches your original)
+        # Linear beta schedule with scaling from second code
         scale = 1000 / timesteps
         beta_start = scale * 0.0001
         beta_end = scale * 0.02
@@ -149,8 +152,7 @@ class DiffusionModel(nn.Module):
         return predicted_noise, noise, t
 
     @torch.no_grad()
-    def sample(self, num_images, img_size, num_classes, labels, device):
-        # Matches your original sampling exactly
+    def sample(self, num_images, img_size, num_classes, labels, device, progress_callback=None):
         x_t = torch.randn(num_images, 3, img_size, img_size).to(device)
 
         if labels.ndim == 1:
@@ -161,6 +163,9 @@ class DiffusionModel(nn.Module):
             labels = labels.to(device)
 
         for t in reversed(range(self.timesteps)):
+            if cancel_event.is_set():
+                return None
+                
             t_tensor = torch.full((num_images,), t, device=device, dtype=torch.float)
             predicted_noise = self.model(x_t, labels, t_tensor)
 
@@ -177,10 +182,13 @@ class DiffusionModel(nn.Module):
                 noise = torch.zeros_like(x_t)
 
             x_t = mean + torch.sqrt(variance) * noise
+            
+            if progress_callback:
+                progress_callback((self.timesteps - t) / self.timesteps)
 
         x_0 = torch.clamp(x_t, -1., 1.)
 
-        # Normalization matching your original code
+        # Normalization
         mean = torch.tensor([0.485, 0.456, 0.406]).view(1, 3, 1, 1).to(device)
         std = torch.tensor([0.229, 0.224, 0.225]).view(1, 3, 1, 1).to(device)
         x_0 = std * x_0 + mean
@@ -188,7 +196,6 @@ class DiffusionModel(nn.Module):
 
         return x_0
 
-# 4. Model Loading (with improved error handling)
 def load_model(model_path, device):
     unet_model = UNet(num_classes=NUM_CLASSES).to(device)
     diffusion_model = DiffusionModel(unet_model, timesteps=TIMESTEPS).to(device)
@@ -228,37 +235,164 @@ def load_model(model_path, device):
     diffusion_model.eval()
     return diffusion_model
 
-# 5. Gradio Interface (matches your original)
-def generate_image(label_str):
+def cancel_generation():
+    cancel_event.set()
+    return "Generation cancelled"
+
+def generate_images(label_str, num_images, progress=gr.Progress()):
+    global loaded_model
+    cancel_event.clear()
+    
+    if num_images < 1 or num_images > 10:
+        raise gr.Error("Number of images must be between 1 and 10")
+    
     label_map = {'Pneumonia': 0, 'Pneumothorax': 1}
     if label_str not in label_map:
-        raise gr.Error("Invalid label selected.")
-
-    label_index = label_map[label_str]
-    labels_to_generate = torch.zeros(1, 2).to(device)
-    labels_to_generate[:, label_index] = 1
-
-    generated_images_tensor = loaded_model.sample(
-        1, IMG_SIZE, NUM_CLASSES, labels_to_generate, device
-    )
-
-    img_np = generated_images_tensor.squeeze(0).permute(1, 2, 0).cpu().numpy()
-    img_pil = Image.fromarray((img_np * 255).astype(np.uint8), 'RGB')
-    return img_pil
-
-# Main Execution
-if __name__ == "__main__":
-    # Load model
-    model_path = "model_weights.pth"  # Match your filename
+        raise gr.Error("Invalid condition selected")
+
+    labels = torch.zeros(num_images, NUM_CLASSES)
+    labels[:, label_map[label_str]] = 1
+
+    try:
+        def progress_callback(progress_val):
+            progress(progress_val, desc="Generating...")
+            if cancel_event.is_set():
+                raise gr.Error("Generation was cancelled by user")
+
+        with torch.no_grad():
+            images = loaded_model.sample(
+                num_images=num_images,
+                img_size=IMG_SIZE,
+                num_classes=NUM_CLASSES,
+                labels=labels,
+                device=device,
+                progress_callback=progress_callback
+            )
+        
+        if images is None:
+            return None, None
+            
+        processed_images = []
+        for img in images:
+            img_np = img.cpu().permute(1, 2, 0).numpy()
+            img_np = (img_np * 255).clip(0, 255).astype(np.uint8)
+            pil_img = Image.fromarray(img_np)
+            processed_images.append(pil_img)
+        
+        if num_images == 1:
+            return processed_images[0], processed_images
+        else:
+            return None, processed_images
+
+    except Exception as e:
+        traceback.print_exc()
+        raise gr.Error(f"Generation failed: {str(e)}")
+    finally:
+        torch.cuda.empty_cache()
+
+# Load model
+MODEL_NAME = "model_weights.pth"
+model_path = MODEL_NAME
+print("Loading model...")
+try:
     loaded_model = load_model(model_path, device)
+    print("Model loaded successfully!")
+except Exception as e:
+    print(f"Failed to load model: {e}")
+    print("Creating dummy model for demonstration")
+    loaded_model = DiffusionModel(UNet(num_classes=NUM_CLASSES), timesteps=TIMESTEPS).to(device)
+
+# Gradio UI (from first code)
+with gr.Blocks(theme=gr.themes.Soft(
+    primary_hue="violet",
+    neutral_hue="slate",
+    font=[gr.themes.GoogleFont("Poppins")],
+    text_size="md"
+)) as demo:
+    gr.Markdown("""
+    <center>
+    <h1>Synthetic X-ray Generator</h1>
+    <p><em>Generate synthetic chest X-rays conditioned on pathology</em></p>
+    </center>
+    """)
+    
+    with gr.Row():
+        with gr.Column(scale=1):
+            condition = gr.Dropdown(
+                ["Pneumonia", "Pneumothorax"],
+                label="Select Condition",
+                value="Pneumonia",
+                interactive=True
+            )
+            num_images = gr.Slider(
+                1, 10, value=1, step=1,
+                label="Number of Images",
+                interactive=True
+            )
+            
+            with gr.Row():
+                submit_btn = gr.Button("Generate", variant="primary")
+                cancel_btn = gr.Button("Cancel", variant="stop")
+            
+            gr.Markdown("""
+            <div style="text-align: center; margin-top: 10px;">
+                <small>Note: Generation may take several seconds per image</small>
+            </div>
+            """)
+        
+        with gr.Column(scale=2):
+            with gr.Tabs():
+                with gr.TabItem("Output", id="output_tab"):
+                    single_image = gr.Image(
+                        label="Generated X-ray",
+                        height=400,
+                        visible=True
+                    )
+                    gallery = gr.Gallery(
+                        label="Generated X-rays",
+                        columns=3,
+                        height="auto",
+                        object_fit="contain",
+                        visible=False
+                    )
     
-    # Create interface
-    iface = gr.Interface(
-        fn=generate_image,
-        inputs=gr.Dropdown(["Pneumonia", "Pneumothorax"], label="Select Condition"),
-        outputs=gr.Image(type="pil", label="Generated X-ray Image"),
-        title="CheXpert X-ray Image Generator",
-        description="Generate synthetic chest X-ray images conditioned on selected conditions (Pneumonia or Pneumothorax) using a diffusion model."
+    def update_ui_based_on_count(num_images):
+        if num_images == 1:
+            return {
+                single_image: gr.update(visible=True),
+                gallery: gr.update(visible=False)
+            }
+        else:
+            return {
+                single_image: gr.update(visible=False),
+                gallery: gr.update(visible=True)
+            }
+    
+    num_images.change(
+        fn=update_ui_based_on_count,
+        inputs=num_images,
+        outputs=[single_image, gallery]
+    )
+    
+    submit_btn.click(
+        fn=generate_images,
+        inputs=[condition, num_images],
+        outputs=[single_image, gallery]
     )
     
-    iface.launch()
+    cancel_btn.click(
+        fn=cancel_generation,
+        outputs=None
+    )
+
+    demo.css = """
+    .gradio-container {
+        background: linear-gradient(135deg, #f5f7fa 0%, #e4e8f0 100%);
+    }
+    .gallery-container {
+        background-color: white !important;
+    }
+    """
+
+if __name__ == "__main__":
+    demo.launch(server_name="0.0.0.0", server_port=7860)