Update app.py

app.py

@@ -27,27 +27,25 @@ def segment_clothing(image):
 
     # Scale back to original size
     mask = cv2.resize(output_predictions, (image.shape[1], image.shape[0]), interpolation=cv2.INTER_NEAREST)
-    print(f"Mask shape: {mask.shape}, unique values: {np.unique(mask)}")  # Debugging
     return mask
 
 def generate_displacement_map(image, mask):
     """Generate a displacement map from the clothing region."""
     gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
-    blurred = cv2.GaussianBlur(gray, (9, 9), 0)  # Reduced kernel size for clarity
+    blurred = cv2.GaussianBlur(gray, (15, 15), 0)
     displacement_map = cv2.normalize(blurred, None, 0, 255, cv2.NORM_MINMAX)
     displacement_map[mask != 15] = 0  # Apply mask (class 15 corresponds to 'person')
-    print(f"Displacement map stats - Min: {np.min(displacement_map)}, Max: {np.max(displacement_map)}")  # Debugging
     return displacement_map
 
 def warp_text(image, text_overlay, displacement_map):
     """Warp the text overlay based on the displacement map."""
     text_overlay_array = np.array(text_overlay)
-    displacement_map = cv2.GaussianBlur(displacement_map, (9, 9), 0)  # Reduced blur for better details
+    displacement_map = cv2.GaussianBlur(displacement_map, (15, 15), 0)
 
     # Create an x, y distortion map
     h, w = displacement_map.shape
     x, y = np.meshgrid(np.arange(w), np.arange(h))
-    x_displacement = x + displacement_map / 100.0  # Adjusted scaling factor for subtle warping
+    x_displacement = x + displacement_map / 100.0
     y_displacement = y + displacement_map / 100.0
 
     # Warp text overlay using remap
@@ -60,57 +58,31 @@ def warp_text(image, text_overlay, displacement_map):
     )
     return Image.fromarray(warped)
 
-def overlay_text(image, text, font_size, color, mask):
+def overlay_text(image, text, font_path, font_size, color, mask, manual_coords=None):
     """Overlay text onto the detected clothing region."""
     pil_image = Image.fromarray(cv2.cvtColor(image, cv2.COLOR_BGR2RGB)).convert("RGBA")
 
-    # Find the bounding box of the mask (clothing area)
-    y_indices, x_indices = np.where(mask == 15)  # Class 15 corresponds to 'person' in DeepLabV3
+    y_indices, x_indices = np.where(mask == 15)
     if len(x_indices) == 0 or len(y_indices) == 0:
         return None, "No clothing region detected."
 
-    x_min, x_max = x_indices.min(), x_indices.max()
-    y_min, y_max = y_indices.min(), y_indices.max()
-
-    clothing_width = x_max - x_min
-    clothing_height = y_max - y_min
-
-    # Ensure the color is correctly formatted
-    color = color.lstrip('#')
-    color_tuple = tuple(int(color[i:i+2], 16) for i in (0, 2, 4))
-
-    # Load font and adjust size dynamically
-    font_path = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"
-    if not os.path.exists(font_path):
-        return None, "Font file not found. Please provide a valid font path."
+    if manual_coords:
+        text_x, text_y = manual_coords
+    else:
+        x_min, x_max = x_indices.min(), x_indices.max()
+        y_min, y_max = y_indices.min(), y_indices.max()
+        text_x, text_y = x_min, y_min
 
+    # Load font and create transparent overlay for text
     font = ImageFont.truetype(font_path, font_size)
-    text_width, text_height = font.getbbox(text)[2:]
-    while text_width > clothing_width or text_height > clothing_height:
-        font_size -= 1
-        if font_size <= 5:
-            return None, "Text too large to fit on the clothing. Try smaller text or font size."
-        font = ImageFont.truetype(font_path, font_size)
-        text_width, text_height = font.getbbox(text)[2:]
-
-    # Calculate position to center the text
-    text_x = x_min + (clothing_width - text_width) // 2
-    text_y = y_min + (clothing_height - text_height) // 2
-
-    # Draw the text on a transparent overlay
     text_overlay = Image.new("RGBA", pil_image.size, (255, 255, 255, 0))
     text_draw = ImageDraw.Draw(text_overlay)
-    try:
-        rgba_color = color_tuple + (255,)  # Add alpha channel
-        text_draw.text((text_x, text_y), text, font=font, fill=rgba_color)
-    except Exception as e:
-        return None, f"Error applying color: {str(e)}"
-
+    text_draw.text((text_x, text_y), text, font=font, fill=color)
     return text_overlay, None
 
-def process_image(image, text, font_size, color):
+def process_image(image, text, font_size, color, font_path, placement_x, placement_y):
     try:
-        # Segment the clothing using DeepLabV3
+        # Segment clothing using DeepLabV3
         mask = segment_clothing(image)
         if mask.sum() == 0:
             return "No clothing detected. Try another image."
@@ -118,33 +90,38 @@ def process_image(image, text, font_size, color):
         # Generate displacement map
         displacement_map = generate_displacement_map(image, mask)
 
-        # Overlay the text
-        text_overlay, error = overlay_text(image, text, font_size, color, mask)
+        # Overlay text with manual coordinates (if provided)
+        manual_coords = (placement_x, placement_y)
+        text_overlay, error = overlay_text(image, text, font_path, font_size, color, mask, manual_coords)
         if error:
             return error
 
         # Warp text using displacement map
         warped_text = warp_text(image, text_overlay, displacement_map)
 
-        # Blend the warped text back onto the original image
+        # Blend warped text with the original image
         pil_image = Image.fromarray(cv2.cvtColor(image, cv2.COLOR_BGR2RGB)).convert("RGBA")
         final_image = Image.alpha_composite(pil_image, warped_text).convert("RGB")
 
         return final_image
     except Exception as e:
-        print(f"Error processing image: {str(e)}")
         return f"Error: {str(e)}"
 
 # Gradio Interface
+font_options = ["/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", "/path/to/another/font.ttf"]
+
 gr.Interface(
     fn=process_image,
     inputs=[
         gr.Image(type="numpy", label="Upload Clothing Image"),
         gr.Textbox(label="Enter Text"),
         gr.Slider(10, 150, step=5, label="Font Size"),
-        gr.ColorPicker(label="Text Color", value="#000000")
+        gr.ColorPicker(label="Text Color", value="#000000"),
+        gr.Dropdown(choices=font_options, label="Select Font", value=font_options[0]),
+        gr.Number(label="Text Placement X", value=0),
+        gr.Number(label="Text Placement Y", value=0),
     ],
     outputs=gr.Image(type="pil", label="Final Image with Warped Text"),
     title="Warped Text Overlay on Clothing",
-    description="Upload a clothing image and add warped text that conforms to folds and curves."
+    description="Upload a clothing image and customize text placement, font, and style."
 ).launch()