Update logic to resize and crop to 1024x1024

app.py

@@ -22,6 +22,7 @@ from addit_methods import add_object_generated, add_object_real
 # Global variables for model
 pipe = None
 device = None
+original_image_size = None
 
 # Initialize model at startup
 print("Initializing ADDIT model...")
@@ -64,6 +65,60 @@ def validate_inputs(prompt_source, prompt_target, subject_token):
         return f"Subject token '{subject_token}' must appear in the target prompt"
     return None
 
+def resize_and_crop_image(image):
+    """
+    Resize and center crop image to 1024x1024.
+    Returns the processed image, a message about what was done, and original size info.
+    """
+    if image is None:
+        return None, "", None
+    
+    original_width, original_height = image.size
+    original_size = (original_width, original_height)
+    
+    # If already 1024x1024, no processing needed
+    if original_width == 1024 and original_height == 1024:
+        return image, "", original_size
+    
+    # Calculate scaling to make smaller dimension 1024
+    scale = 1024 / min(original_width, original_height)
+    new_width = int(original_width * scale)
+    new_height = int(original_height * scale)
+    
+    # Resize image
+    resized_image = image.resize((new_width, new_height), Image.Resampling.LANCZOS)
+    
+    # Center crop to 1024x1024
+    left = (new_width - 1024) // 2
+    top = (new_height - 1024) // 2
+    right = left + 1024
+    bottom = top + 1024
+    
+    cropped_image = resized_image.crop((left, top, right, bottom))
+    
+    # Create status message
+    if new_width == 1024 and new_height == 1024:
+        message = f"<div style='background-color: #e8f5e8; border: 1px solid #4caf50; border-radius: 5px; padding: 8px; margin-bottom: 10px;'><span style='color: #2e7d32; font-weight: bold;'>✅ Image resized to 1024×1024</span></div>"
+    else:
+        message = f"<div style='background-color: #e8f5e8; border: 1px solid #4caf50; border-radius: 5px; padding: 8px; margin-bottom: 10px;'><span style='color: #2e7d32; font-weight: bold;'>✅ Image resized and center cropped to 1024×1024</span></div>"
+    
+    return cropped_image, message, original_size
+
+def handle_image_upload(image):
+    """Handle image upload and store original size globally"""
+    global original_image_size
+    
+    if image is None:
+        original_image_size = None
+        return None, ""
+    
+    # Store original size
+    original_image_size = image.size
+    
+    # Process image
+    processed_image, message, _ = resize_and_crop_image(image)
+    return processed_image, message
+
 @spaces.GPU
 def process_generated_image(
     prompt_source,
@@ -161,6 +216,8 @@ def process_real_image(
     # Print current time and input information
     current_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
     print(f"\n[{current_time}] Starting Real Image Processing")
+    if original_image_size:
+        print(f"Original uploaded image size: {original_image_size[0]}×{original_image_size[1]}")
     print(f"Source Image Size: {source_image.size}")
     print(f"Source Prompt: '{prompt_source}'")
     print(f"Target Prompt: '{prompt_target}'")
@@ -319,10 +376,11 @@ def create_interface():
             # Real Images Tab
             with gr.TabItem("📸 Real Images"):
                 gr.Markdown("### Upload an image and add objects to it")
-                gr.HTML("<p style='color: red; font-weight: bold; margin: -15px -10px;'>Note: Images will be resized to 1024x1024 pixels. For best results, use square images.</p>")
+                gr.HTML("<p style='color: orange; font-weight: bold; margin: -15px -10px;'>Note: Images will be automatically resized and center cropped to 1024×1024 pixels.</p>")
                 
                 with gr.Row():
                     with gr.Column(scale=1):
+                        real_image_status = gr.HTML(visible=False)
                         real_source_image = gr.Image(label="Source Image", type="pil")
                         real_prompt_source = gr.Textbox(
                             label="Source Prompt",
@@ -385,6 +443,17 @@ def create_interface():
                             real_edited_output = gr.Image(label="Edited Image", type="pil")
                         real_status = gr.Textbox(label="Status", interactive=False)
                 
+                # Handle image upload and preprocessing
+                real_source_image.upload(
+                    fn=handle_image_upload,
+                    inputs=[real_source_image],
+                    outputs=[real_source_image, real_image_status]
+                ).then(
+                    fn=lambda status: gr.update(visible=bool(status.strip()), value=status),
+                    inputs=[real_image_status],
+                    outputs=[real_image_status]
+                )
+                
                 real_submit_btn.click(
                     fn=process_real_image,
                     inputs=[