Update app.py

app.py

@@ -1,5 +1,3 @@
-#!/usr/bin/env python
-
 import os
 import random
 import uuid
@@ -11,69 +9,13 @@ from PIL import Image
 import spaces
 import torch
 from diffusers import DiffusionPipeline
+import face_recognition  # More robust face detection library
 from typing import Tuple
 
-# Setup rules for bad words (ensure the prompts are kid-friendly)
-bad_words = json.loads(os.getenv('BAD_WORDS', '["violence", "blood", "scary", "death", "ghost"]'))
-default_negative = os.getenv("default_negative","")
-
-def check_text(prompt, negative=""):
-    for i in bad_words:
-        if i in prompt:
-            return True
-    return False
-
-# Kid-friendly styles
-style_list = [
-    {
-        "name": "Cartoon",
-        "prompt": "colorful cartoon {prompt}. vibrant, playful, friendly, suitable for children, highly detailed, bright colors",
-        "negative_prompt": "scary, dark, violent, ugly, realistic",
-    },
-    {
-        "name": "Children's Illustration",
-        "prompt": "children's illustration {prompt}. cute, colorful, fun, simple shapes, smooth lines, highly detailed, joyful",
-        "negative_prompt": "scary, dark, violent, deformed, ugly",
-    },   
-    {
-        "name": "Sticker",
-        "prompt": "children's sticker of {prompt}. bright colors, playful, high resolution, cartoonish",
-        "negative_prompt": "scary, dark, violent, ugly, low resolution",
-    },
-    {
-        "name": "Fantasy",
-        "prompt": "fantasy world for children with {prompt}. magical, vibrant, friendly, beautiful, colorful",
-        "negative_prompt": "dark, scary, violent, ugly, realistic",
-    },
-    {
-        "name": "(No style)",
-        "prompt": "{prompt}",
-        "negative_prompt": "",
-    },
-]
-
-styles = {k["name"]: (k["prompt"], k["negative_prompt"]) for k in style_list}
-STYLE_NAMES = list(styles.keys())
-DEFAULT_STYLE_NAME = "Sticker"
-
-def apply_style(style_name: str, positive: str, negative: str = "") -> Tuple[str, str]:
-    p, n = styles.get(style_name, styles[DEFAULT_STYLE_NAME])
-    return p.replace("{prompt}", positive), n + negative
-
-DESCRIPTION = """## Children's Sticker Generator
-
-Generate fun and playful stickers for children using AI.
-"""
-
-if not torch.cuda.is_available():
-    DESCRIPTION += "\n<p>⚠️Running on CPU, This may not work on CPU.</p>"
-
-MAX_SEED = np.iinfo(np.int32).max
-CACHE_EXAMPLES = torch.cuda.is_available() and os.getenv("CACHE_EXAMPLES", "0") == "1"
-
+# Check if GPU is available; fallback to CPU if needed
 device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
-# Initialize the DiffusionPipeline
+# Initialize the AI model for sticker generation
 pipe = DiffusionPipeline.from_pretrained(
     "SG161222/RealVisXL_V3.0_Turbo",  # or any model of your choice
     torch_dtype=torch.float16,
@@ -81,193 +23,124 @@ pipe = DiffusionPipeline.from_pretrained(
     variant="fp16"
 ).to(device)
 
-# Convert mm to pixels for a specific DPI (300) and ensure divisible by 8
-def mm_to_pixels(mm, dpi=300):
-    """Convert mm to pixels and make the dimensions divisible by 8."""
-    pixels = int((mm / 25.4) * dpi)
-    return pixels - (pixels % 8)  # Adjust to the nearest lower multiple of 8
 
-# Default sizes for 75mm and 35mm, rounded to nearest multiple of 8
-size_map = {
-    "75mm": (mm_to_pixels(75), mm_to_pixels(75)),  # 75mm in pixels at 300dpi
-    "35mm": (mm_to_pixels(35), mm_to_pixels(35)),  # 35mm in pixels at 300dpi
-}
+def face_to_sticker(image_path: str) -> Tuple[str, str]:
+    """Detect the face using face_recognition and convert it to a sticker format."""
+    img = face_recognition.load_image_file(image_path)
+    face_locations = face_recognition.face_locations(img)
 
-# Function to post-process images (transparent or white background)
-def save_image(img, background="transparent"):
-    img = img.convert("RGBA")
-    data = img.getdata()
-    new_data = []
+    if not face_locations:
+        return None, "No face detected. Please upload a clear image with a visible face."
 
-    if background == "transparent":
-        for item in data:
-            # Replace white with transparent
-            if item[0] == 255 and item[1] == 255 and item[2] == 255:
-                new_data.append((255, 255, 255, 0))  # Transparent
-            else:
-                new_data.append(item)
-    elif background == "white":
-        for item in data:
-            new_data.append(item)  # Keep as white
-
-    img.putdata(new_data)
-    unique_name = str(uuid.uuid4()) + ".png"
-    img.save(unique_name)
-    return unique_name
-
-def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
-    if randomize_seed:
-        seed = random.randint(0, MAX_SEED)
-    return seed
-
-@spaces.GPU(enable_queue=True)
-def generate(
-    prompt: str,
-    negative_prompt: str = "",
-    use_negative_prompt: bool = False,
-    style: str = DEFAULT_STYLE_NAME,
-    seed: int = 0,
-    size: str = "75mm",
-    guidance_scale: float = 3,
-    randomize_seed: bool = False,
-    background: str = "transparent",
-    progress=gr.Progress(track_tqdm=True),
-):
-    if check_text(prompt, negative_prompt):
-        raise ValueError("Prompt contains restricted words.")
+    # Extract the first detected face and return as an image for sticker creation
+    top, right, bottom, left = face_locations[0]
+    face_img = img[top:bottom, left:right]
+    face_img = Image.fromarray(face_img).resize((256, 256))  # Resize face to sticker size
     
-    # Ensure prompt is 2-3 words long
-    prompt = " ".join(re.findall(r'\w+', prompt)[:3])
+    face_img_path = f"{uuid.uuid4()}.png"
+    face_img.save(face_img_path)
+    return face_img_path, "Face successfully converted to a sticker."
 
-    # Apply style
-    prompt, negative_prompt = apply_style(style, prompt, negative_prompt)
-    seed = int(randomize_seed_fn(seed, randomize_seed))
-    generator = torch.Generator().manual_seed(seed)
 
-    # Ensure we have only white or transparent background options
-    width, height = size_map.get(size, (1024, 1024))
+def generate_prompt(clothing: str, pose: str, mood: str) -> str:
+    """Generate a descriptive prompt based on user-selected clothing, pose, and mood."""
+    prompt = f"sticker of a person wearing {clothing} clothes, in a {pose} pose, looking {mood}."
+    return prompt
 
-    if not use_negative_prompt:
-        negative_prompt = ""  # type: ignore
 
+def generate_stickers(prompt: str, face_image: str, guidance_scale: float = 7.5, randomize_seed: bool = False):
+    """Generate stickers using the diffusion model with the given prompt and face."""
+    
+    # Adjust seed for variability
+    seed = random.randint(0, MAX_SEED) if randomize_seed else 42
+    generator = torch.Generator(device).manual_seed(seed)
+
+    # Prepare AI model options
     options = {
         "prompt": prompt,
-        "negative_prompt": negative_prompt,
-        "width": width,
-        "height": height,
+        "width": 512,
+        "height": 512,
         "guidance_scale": guidance_scale,
-        "num_inference_steps": 25,
+        "num_inference_steps": 50,
         "generator": generator,
-        "num_images_per_prompt": 6,  # Max 6 images
-        "output_type": "pil",
     }
-    
-    # Generate images with the pipeline
+
+    # Load the face as an input condition for the sticker (optional, if supported by the model)
+    # If your model supports conditioning on a specific face, load the face image here
+    # options['image'] = Image.open(face_image)
+
     images = pipe(**options).images
-    image_paths = [save_image(img, background) for img in images]
+    image_paths = [save_image(img) for img in images]
     
     return image_paths, seed
 
-examples = [
-    "cute bunny",
-    "happy cat",
-    "funny dog",
-]
 
-css = '''
-.gradio-container{max-width: 700px !important}
-h1{text-align:center}
-'''
+def save_image(img: Image.Image) -> str:
+    """Save an image to a file and return the path."""
+    unique_name = f"{uuid.uuid4()}.png"
+    img.save(unique_name)
+    return unique_name
 
-# Define the Gradio UI for the sticker generator
-with gr.Blocks(css=css, theme="bethecloud/storj_theme") as demo:
-    gr.Markdown(DESCRIPTION)
-    gr.DuplicateButton(
-        value="Duplicate Space for private use",
-        elem_id="duplicate-button",
-        visible=os.getenv("SHOW_DUPLICATE_BUTTON") == "1",
-    )
-    with gr.Group():
-        with gr.Row():
-            prompt = gr.Text(
-                label="Enter your prompt",
-                show_label=False,
-                max_lines=1,
-                placeholder="Enter 2-3 word prompt (e.g., cute bunny)",
-                container=False,
-            )
-            run_button = gr.Button("Run")
-        result = gr.Gallery(label="Generated Stickers", columns=2, preview=True)
-    with gr.Accordion("Advanced options", open=False):
-        use_negative_prompt = gr.Checkbox(label="Use negative prompt", value=True, visible=True)
-        negative_prompt = gr.Text(
-            label="Negative prompt",
-            max_lines=1,
-            placeholder="Enter a negative prompt",
-            value="(scary, violent, dark, ugly)",
-            visible=True,
-        )
-        seed = gr.Slider(
-            label="Seed",
-            minimum=0,
-            maximum=MAX_SEED,
-            step=1,
-            value=0,
-            visible=True
-        )
-        randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
-        size_selection = gr.Radio(
-            choices=["75mm", "35mm"],
-            value="75mm",
-            label="Sticker Size",
-        )
-        style_selection = gr.Radio(
-            choices=STYLE_NAMES,
-            value=DEFAULT_STYLE_NAME,
-            label="Image Style",
-        )
-        background_selection = gr.Radio(
-            choices=["transparent", "white"],
-            value="transparent",
-            label="Background Color",
-        )
-        guidance_scale = gr.Slider(
-            label="Guidance Scale",
-            minimum=0.1,
-            maximum=20.0,
-            step=0.1,
-            value=6,
-        )
 
-    gr.Examples(
-        examples=examples,
-        inputs=prompt,
-        outputs=[result, seed],
-        fn=generate,
-        cache_examples=CACHE_EXAMPLES,
+def stick_me_workflow(image, clothing, pose, mood, randomize_seed: bool):
+    """Workflow to generate stickers based on user-uploaded image and options."""
+    # Convert the uploaded image to a face sticker
+    face_path, message = face_to_sticker(image)
+    
+    if face_path is None:
+        return message  # Return error message if face detection fails
+
+    # Generate a descriptive prompt based on user selections
+    prompt = generate_prompt(clothing, pose, mood)
+    
+    # Generate stickers using the diffusion model with the extracted face and prompt
+    stickers, seed = generate_stickers(prompt, face_path, randomize_seed=randomize_seed)
+    return stickers
+
+
+def on_fallback_to_cpu():
+    """Notify users when the app is running on CPU (due to GPU quota being exceeded)."""
+    if not torch.cuda.is_available():
+        return "Warning: GPU quota exceeded. Running on CPU, which will be significantly slower."
+    return ""
+
+
+# Gradio interface setup
+
+with gr.Blocks() as demo:
+    gr.Markdown("# Sticker Generator with 'Stick Me' Feature")
+
+    # GPU Quota Handling
+    gpu_warning = gr.Markdown(on_fallback_to_cpu(), visible=not torch.cuda.is_available())
+
+    # New Stick Me Option
+    with gr.Row():
+        face_input = gr.Image(label="Upload Your Image for 'Stick Me'", type="filepath")
+        clothing = gr.Dropdown(["Casual", "Formal", "Sports"], label="Choose Clothing")
+        pose = gr.Dropdown(["Standing", "Sitting", "Running"], label="Choose Pose")
+        mood = gr.Dropdown(["Happy", "Serious", "Excited"], label="Choose Mood")
+        randomize_seed = gr.Checkbox(label="Randomize Seed", value=True)
+    
+        stick_me_button = gr.Button("Generate Stick Me Stickers")
+    
+    stick_me_result = gr.Gallery(label="Your Stick Me Stickers")
+    
+    stick_me_button.click(
+        fn=stick_me_workflow,
+        inputs=[face_input, clothing, pose, mood, randomize_seed],
+        outputs=[stick_me_result]
     )
+    
+    gr.Markdown("# Generate Regular Stickers")
+
+    prompt = gr.Textbox(label="Enter a Prompt for Sticker Creation", placeholder="Cute bunny", max_lines=1)
+    generate_button = gr.Button("Generate Stickers")
+    result = gr.Gallery(label="Generated Stickers")
 
-    gr.on(
-        triggers=[
-            prompt.submit,
-            negative_prompt.submit,
-            run_button.click,
-        ],
-        fn=generate,
-        inputs=[
-            prompt,
-            negative_prompt,
-            use_negative_prompt,
-            style_selection,
-            seed,
-            size_selection,
-            guidance_scale,
-            randomize_seed,
-            background_selection,
-        ],
-        outputs=[result, seed],
-        api_name="run",
+    generate_button.click(
+        fn=generate_stickers,
+        inputs=[prompt],
+        outputs=[result]
     )
 
-if __name__ == "__main__":
-    demo.queue(max_size=20).launch()
+    demo.launch()