Added application file

app.py

@@ -1,31 +1,124 @@
 import gradio as gr
-import time
+import numpy as np
+from PIL import Image
+import tempfile
+import os
 
 
-def sleep(im):
-    time.sleep(5)
-    return [im["background"], im["layers"][0], im["layers"][1], im["composite"]]
+# Function to create the binary mask from the ImageEditor's output
+def create_binary_mask(im_dict):
+    if im_dict is None or im_dict["background"] is None:
+        print("No background image found.")
+        # Return a small blank placeholder and None for the file path
+        blank_preview = np.zeros((768, 1024), dtype=np.uint8)
+        return blank_preview, None
 
+    background_img = im_dict["background"]
+    h, w, _ = background_img.shape # Get original dimensions (Height, Width, Channels)
+    print(f"Original image dimensions: H={h}, W={w}")
 
-def predict(im):
-    return im["composite"]
+    # Check if any drawing layer exists and is not None
+    if not im_dict["layers"] or im_dict["layers"][0] is None:
+        print("No drawing layer found. Generating blank mask.")
+        # Nothing drawn yet, return a black mask of the original size
+        mask = np.zeros((h, w), dtype=np.uint8)
+        filepath = None # No file to download as nothing was drawn
+    else:
+        # Use the first layer (index 0) which usually contains the drawing
+        layer = im_dict["layers"][0]
+        print(f"Drawing layer dimensions: H={layer.shape[0]}, W={layer.shape[1]}")
 
+        # Ensure layer dimensions match background (Gradio ImageEditor usually handles this)
+        if layer.shape[0] != h or layer.shape[1] != w:
+            print(f"Warning: Layer size ({layer.shape[0]}x{layer.shape[1]}) doesn't match background ({h}x{w}). This shouldn't happen.")
+            # Handle potential mismatch if necessary, though unlikely with default editor behavior
+            # For now, proceed assuming they match or the layer is the correct reference
 
+        # Layer is RGBA, extract the Alpha channel (index 3)
+        alpha_channel = layer[:, :, 3]
+
+        # Create binary mask: white (255) where alpha > 0 (drawn), black (0) otherwise
+        mask = np.where(alpha_channel > 0, 255, 0).astype(np.uint8)
+        print(f"Generated binary mask dimensions: H={mask.shape[0]}, W={mask.shape[1]}")
+
+        # Save the mask to a temporary PNG file for download
+        try:
+            # Create a temporary file path
+            with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmpfile:
+                filepath = tmpfile.name
+            
+            # Save the NumPy array as a PNG image using PIL
+            pil_image = Image.fromarray(mask)
+            pil_image.save(filepath, format="PNG")
+            print(f"Mask saved temporarily to: {filepath}")
+
+        except Exception as e:
+            print(f"Error saving mask to temporary file: {e}")
+            filepath = None # Indicate failure to save
+            # Return a blank mask in case of saving error
+            mask = np.zeros((h, w), dtype=np.uint8)
+
+    # Return the mask NumPy array for preview and the filepath for download
+    # The DownloadButton component will become active/functional if filepath is not None
+    return mask, filepath
+
+# --- Gradio App Layout ---
 with gr.Blocks() as demo:
+    gr.Markdown("## Binary Mask Generator")
+    gr.Markdown(
+        "Upload or paste an image. Use the brush tool (select it!) to draw the area "
+        "you want to mask. Click 'Generate Mask' to see the result and download it."
+    )
+
     with gr.Row():
-        im = gr.ImageEditor(
-            type="numpy",
-            crop_size="1:1",
-        )
-        im_preview = gr.Image()
-    n_upload = gr.Number(0, label="Number of upload events", step=1)
-    n_change = gr.Number(0, label="Number of change events", step=1)
-    n_input = gr.Number(0, label="Number of input events", step=1)
-
-    im.upload(lambda x: x + 1, outputs=n_upload, inputs=n_upload)
-    im.change(lambda x: x + 1, outputs=n_change, inputs=n_change)
-    im.input(lambda x: x + 1, outputs=n_input, inputs=n_input)
-    im.change(predict, outputs=im_preview, inputs=im, show_progress="hidden")
+        # --- Left Column ---
+        with gr.Column(scale=1): # Adjust scale as needed
+            image_editor = gr.ImageEditor(
+                label="Draw on Image",
+                # type="numpy" is essential for processing layers
+                type="numpy",
+                # DON'T set crop_size, height, or width to keep original dimensions
+                # sources allow upload, paste, webcam etc.
+                sources=["upload"],
+                # Set a default brush for clarity (optional, but helpful)
+                brush=gr.Brush(colors=["#FF0000"], color_mode="fixed"), # Red fixed brush
+                interactive=True,
+                canvas_size=(768, 1024)
+            )
+            generate_button = gr.Button("Generate Mask", variant="primary")
+
+        # --- Right Column ---
+        with gr.Column(scale=1): # Adjust scale as needed
+            mask_preview = gr.Image(
+                label="Binary Mask Preview",
+                # Use numpy for consistency, PIL would also work
+                type="numpy",
+                interactive=False, # Preview is not interactive
+            )
+            # Download button - its value (the file path) is set by the function's output
+            download_button = gr.DownloadButton(
+                label="Download Mask (PNG)",
+                interactive=True, # Button starts interactive
+            )
 
+    # --- Event Handling ---
+    generate_button.click(
+        fn=create_binary_mask,
+        inputs=[image_editor],
+        # Output 1 goes to mask_preview (image data)
+        # Output 2 goes to download_button (file path for the 'value' argument)
+        outputs=[mask_preview, download_button]
+    )
+
+# --- Launch the App ---
 if __name__ == "__main__":
-    demo.launch()
+    # Cleaning up old temp files on startup (optional but good practice)
+    temp_dir = tempfile.gettempdir()
+    for item in os.listdir(temp_dir):
+        if item.endswith(".png") and item.startswith("tmp"): # Be specific to avoid deleting wrong files
+             try:
+                 os.remove(os.path.join(temp_dir, item))
+             except Exception:
+                 pass # Ignore if file is locked etc.
+
+    demo.launch(share=True)

old_app.py

@@ -1,135 +0,0 @@
-import gradio as gr
-import numpy as np
-from PIL import Image
-import tempfile
-import os
-
-# Function to create the binary mask from the ImageEditor's output
-def create_binary_mask(im_dict):
-    """
-    Generates a binary mask from the drawing layer of the gr.ImageEditor output.
-
-    Args:
-        im_dict (dict): The dictionary output from gr.ImageEditor, containing
-                        'background', 'layers', and 'composite'.
-
-    Returns:
-        tuple: A tuple containing:
-            - np.ndarray: The binary mask image (H, W) as a NumPy array (0 or 255).
-            - str or None: The filepath to the saved PNG mask for download, or None if no mask generated.
-    """
-    if im_dict is None or im_dict["background"] is None:
-        print("No background image found.")
-        # Return a small blank placeholder and None for the file path
-        blank_preview = np.zeros((768, 1024), dtype=np.uint8)
-        return blank_preview, None
-
-    background_img = im_dict["background"]
-    h, w, _ = background_img.shape # Get original dimensions (Height, Width, Channels)
-    print(f"Original image dimensions: H={h}, W={w}")
-
-    # Check if any drawing layer exists and is not None
-    if not im_dict["layers"] or im_dict["layers"][0] is None:
-        print("No drawing layer found. Generating blank mask.")
-        # Nothing drawn yet, return a black mask of the original size
-        mask = np.zeros((h, w), dtype=np.uint8)
-        filepath = None # No file to download as nothing was drawn
-    else:
-        # Use the first layer (index 0) which usually contains the drawing
-        layer = im_dict["layers"][0]
-        print(f"Drawing layer dimensions: H={layer.shape[0]}, W={layer.shape[1]}")
-
-        # Ensure layer dimensions match background (Gradio ImageEditor usually handles this)
-        if layer.shape[0] != h or layer.shape[1] != w:
-            print(f"Warning: Layer size ({layer.shape[0]}x{layer.shape[1]}) doesn't match background ({h}x{w}). This shouldn't happen.")
-            # Handle potential mismatch if necessary, though unlikely with default editor behavior
-            # For now, proceed assuming they match or the layer is the correct reference
-
-        # Layer is RGBA, extract the Alpha channel (index 3)
-        alpha_channel = layer[:, :, 3]
-
-        # Create binary mask: white (255) where alpha > 0 (drawn), black (0) otherwise
-        mask = np.where(alpha_channel > 0, 255, 0).astype(np.uint8)
-        print(f"Generated binary mask dimensions: H={mask.shape[0]}, W={mask.shape[1]}")
-
-        # Save the mask to a temporary PNG file for download
-        try:
-            # Create a temporary file path
-            with tempfile.NamedTemporaryFile(suffix=".png", delete=False) as tmpfile:
-                filepath = tmpfile.name
-            
-            # Save the NumPy array as a PNG image using PIL
-            pil_image = Image.fromarray(mask)
-            pil_image.save(filepath, format="PNG")
-            print(f"Mask saved temporarily to: {filepath}")
-
-        except Exception as e:
-            print(f"Error saving mask to temporary file: {e}")
-            filepath = None # Indicate failure to save
-            # Return a blank mask in case of saving error
-            mask = np.zeros((h, w), dtype=np.uint8)
-
-    # Return the mask NumPy array for preview and the filepath for download
-    # The DownloadButton component will become active/functional if filepath is not None
-    return mask, filepath
-
-# --- Gradio App Layout ---
-with gr.Blocks() as demo:
-    gr.Markdown("## Binary Mask Generator")
-    gr.Markdown(
-        "Upload or paste an image. Use the brush tool (select it!) to draw the area "
-        "you want to mask. Click 'Generate Mask' to see the result and download it."
-    )
-
-    with gr.Row():
-        # --- Left Column ---
-        with gr.Column(scale=1): # Adjust scale as needed
-            image_editor = gr.ImageEditor(
-                label="Draw on Image",
-                # type="numpy" is essential for processing layers
-                type="numpy",
-                # DON'T set crop_size, height, or width to keep original dimensions
-                # sources allow upload, paste, webcam etc.
-                sources=["upload"],
-                # Set a default brush for clarity (optional, but helpful)
-                brush=gr.Brush(colors=["#FF0000"], color_mode="fixed"), # Red fixed brush
-                interactive=True,
-                canvas_size=(768, 1024)
-            )
-            generate_button = gr.Button("Generate Mask", variant="primary")
-
-        # --- Right Column ---
-        with gr.Column(scale=1): # Adjust scale as needed
-            mask_preview = gr.Image(
-                label="Binary Mask Preview",
-                # Use numpy for consistency, PIL would also work
-                type="numpy",
-                interactive=False, # Preview is not interactive
-            )
-            # Download button - its value (the file path) is set by the function's output
-            download_button = gr.DownloadButton(
-                label="Download Mask (PNG)",
-                interactive=True, # Button starts interactive
-            )
-
-    # --- Event Handling ---
-    generate_button.click(
-        fn=create_binary_mask,
-        inputs=[image_editor],
-        # Output 1 goes to mask_preview (image data)
-        # Output 2 goes to download_button (file path for the 'value' argument)
-        outputs=[mask_preview, download_button]
-    )
-
-# --- Launch the App ---
-if __name__ == "__main__":
-    # Cleaning up old temp files on startup (optional but good practice)
-    temp_dir = tempfile.gettempdir()
-    for item in os.listdir(temp_dir):
-        if item.endswith(".png") and item.startswith("tmp"): # Be specific to avoid deleting wrong files
-             try:
-                 os.remove(os.path.join(temp_dir, item))
-             except Exception:
-                 pass # Ignore if file is locked etc.
-
-    demo.launch()

requirements.txt

@@ -12,5 +12,5 @@ fvcore==0.1.5.post20221221
 cloudpickle==3.0.0
 omegaconf==2.3.0
 av==12.3.0
-gradio==5.0.1
+gradio==4.41.*
 huggingface_hub==0.25.0