Spaces:
Running
Running
| #!/usr/bin/env python3 | |
| """ | |
| Affecto - Real MagicFace Emotion Transformation | |
| Made by Gaurav Jha and Raj Shakya | |
| """ | |
| # ============================================ | |
| # CRITICAL: Fix huggingface-hub version FIRST | |
| # ============================================ | |
| import subprocess | |
| import sys | |
| import os | |
| print("=" * 60) | |
| print("🔧 FIXING DEPENDENCY VERSIONS") | |
| print("=" * 60) | |
| # Check current version | |
| try: | |
| import huggingface_hub | |
| current_version = huggingface_hub.__version__ | |
| print(f"📦 Current huggingface-hub version: {current_version}") | |
| if current_version != "0.25.2": | |
| print(f"❌ Version {current_version} doesn't match required 0.25.2!") | |
| print("🔄 Force installing 0.25.2...") | |
| # Uninstall current version | |
| subprocess.check_call([ | |
| sys.executable, "-m", "pip", "uninstall", | |
| "huggingface-hub", "-y" | |
| ], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) | |
| # Install correct version | |
| subprocess.check_call([ | |
| sys.executable, "-m", "pip", "install", | |
| "huggingface-hub==0.25.2", "--no-deps" | |
| ], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL) | |
| print(f"✅ Installed huggingface-hub 0.25.2") | |
| else: | |
| print(f"✅ Version {current_version} is correct!") | |
| except Exception as e: | |
| print(f"⚠️ Error fixing huggingface-hub: {str(e)}") | |
| print("⚠️ Proceeding anyway...") | |
| print("=" * 60) | |
| print() | |
| # ============================================ | |
| # HF Hub compatibility shim | |
| # ============================================ | |
| import importlib | |
| import importlib.util | |
| hf_spec = importlib.util.find_spec("huggingface_hub") | |
| if hf_spec is not None: | |
| hf = importlib.import_module("huggingface_hub") | |
| if not hasattr(hf, "cached_download") and hasattr(hf, "hf_hub_download"): | |
| def cached_download(*args, **kwargs): | |
| return hf.hf_hub_download(*args, **kwargs) | |
| setattr(hf, "cached_download", cached_download) | |
| print("shim: huggingface_hub cached_download patched:", hasattr(hf, "cached_download")) | |
| else: | |
| print("shim: huggingface_hub not present at import-time") | |
| # ============================================ | |
| # NOW SAFE TO IMPORT OTHER MODULES | |
| # ============================================ | |
| import gradio as gr | |
| import torch | |
| from PIL import Image | |
| import base64 | |
| from io import BytesIO | |
| import json | |
| import traceback | |
| print("🚀 Starting Affecto - Real MagicFace Inference Service...") | |
| device = "cuda" if torch.cuda.is_available() else "cpu" | |
| print(f"🖥️ Device: {device}") | |
| # Initialize preprocessor and model | |
| print("📥 Loading models...") | |
| from preprocessor import FacePreprocessor | |
| from magicface_model import MagicFaceModel | |
| try: | |
| preprocessor = FacePreprocessor(device=device) | |
| model = MagicFaceModel(device=device) | |
| print("✅ All models loaded successfully!") | |
| except Exception as e: | |
| print(f"❌ Error loading models: {str(e)}") | |
| traceback.print_exc() | |
| raise | |
| # ============================================ | |
| # UTILITY FUNCTIONS | |
| # ============================================ | |
| def pil_to_base64(image): | |
| """Convert PIL to base64""" | |
| buffered = BytesIO() | |
| image.save(buffered, format="JPEG", quality=95) | |
| return base64.b64encode(buffered.getvalue()).decode() | |
| def base64_to_pil(base64_str): | |
| """Convert base64 to PIL""" | |
| if base64_str.startswith('data:image'): | |
| base64_str = base64_str.split(',')[1] | |
| image_bytes = base64.b64decode(base64_str) | |
| return Image.open(BytesIO(image_bytes)).convert('RGB') | |
| # ============================================ | |
| # INFERENCE FUNCTION | |
| # ============================================ | |
| def transform_emotion(image, au_params_str, steps, seed): | |
| """ | |
| Main transformation function | |
| Args: | |
| image: PIL Image | |
| au_params_str: JSON string of AU parameters | |
| steps: Number of inference steps | |
| seed: Random seed | |
| Returns: | |
| result_image: Transformed PIL Image | |
| status_msg: Status message | |
| """ | |
| try: | |
| if image is None: | |
| return None, "❌ No image provided" | |
| print("\n" + "="*60) | |
| print("🎭 NEW TRANSFORMATION REQUEST") | |
| print("="*60) | |
| # Parse AU params | |
| try: | |
| au_params = json.loads(au_params_str) | |
| print(f"✅ AU Parameters: {au_params}") | |
| except json.JSONDecodeError as e: | |
| return None, f"❌ Invalid JSON: {str(e)}" | |
| # Step 1: Preprocess (detect face, crop, extract background) | |
| print("\n📸 STEP 1: Preprocessing...") | |
| try: | |
| source_img, bg_img = preprocessor.preprocess(image) | |
| print("✅ Preprocessing complete") | |
| except Exception as e: | |
| return None, f"❌ Preprocessing failed: {str(e)}" | |
| # Step 2: Transform with MagicFace | |
| print("\n🎨 STEP 2: MagicFace Transformation...") | |
| print(f" Inference steps: {steps}") | |
| print(f" Seed: {seed}") | |
| print(f" Expected time: ~{steps * 10} seconds on CPU") | |
| try: | |
| result_img = model.transform( | |
| source_image=source_img, | |
| bg_image=bg_img, | |
| au_params=au_params, | |
| num_inference_steps=steps, | |
| seed=seed | |
| ) | |
| print("\n✅ TRANSFORMATION SUCCESSFUL!") | |
| print("="*60 + "\n") | |
| return result_img, "✅ Transformation successful!" | |
| except Exception as e: | |
| print(f"\n❌ Transformation failed: {str(e)}") | |
| traceback.print_exc() | |
| return None, f"❌ Transformation failed: {str(e)}" | |
| except Exception as e: | |
| print(f"\n❌ Unexpected error: {str(e)}") | |
| traceback.print_exc() | |
| return None, f"❌ Error: {str(e)}" | |
| # ============================================ | |
| # GRADIO INTERFACE | |
| # ============================================ | |
| with gr.Blocks(theme=gr.themes.Soft(), title="Affecto - Real MagicFace") as demo: | |
| gr.Markdown("# 🎭 Affecto - Real MagicFace Emotion Transformation") | |
| gr.Markdown("Transform facial emotions using Action Units with the complete MagicFace diffusion model") | |
| gr.Markdown(f"⚡ Running on: **{device.upper()}** | ⏱️ Estimated time: ~10-15 minutes per transformation on CPU") | |
| with gr.Tab("🖼️ Transform"): | |
| with gr.Row(): | |
| with gr.Column(): | |
| input_image = gr.Image(type="pil", label="Upload Face Image") | |
| gr.Markdown("### 🎛️ Action Unit Parameters") | |
| au_params_input = gr.Textbox( | |
| label="AU Parameters (JSON)", | |
| value='{"AU6": 2.0, "AU12": 2.0}', | |
| lines=4, | |
| placeholder='{"AU6": 2.0, "AU12": 2.0}' | |
| ) | |
| with gr.Row(): | |
| steps_slider = gr.Slider( | |
| minimum=20, maximum=100, value=50, step=10, | |
| label="Inference Steps (20=fast/lower quality, 50=balanced, 100=slow/high quality)" | |
| ) | |
| with gr.Row(): | |
| seed_input = gr.Number(value=424, label="Seed (for reproducibility)", precision=0) | |
| transform_btn = gr.Button("✨ Transform Emotion", variant="primary", size="lg") | |
| gr.Markdown("⚠️ **Note:** Transformation takes ~10-15 minutes on CPU. Please be patient!") | |
| with gr.Column(): | |
| output_image = gr.Image(type="pil", label="Transformed Result") | |
| status_text = gr.Textbox(label="Status", interactive=False, lines=2) | |
| gr.Markdown("### 🎨 Emotion Presets (click to use):") | |
| gr.Examples( | |
| examples=[ | |
| ['{"AU6": 2.5, "AU12": 3.0}', 50, 424], # Happy | |
| ['{"AU1": 2.0, "AU4": 2.5, "AU15": 2.5}', 50, 424], # Sad | |
| ['{"AU4": 3.0, "AU5": 2.0, "AU7": 2.5, "AU23": 2.0}', 50, 424], # Angry | |
| ['{"AU1": 3.0, "AU2": 2.5, "AU5": 3.0, "AU26": 2.5}', 50, 424], # Surprised | |
| ['{"AU6": 1.5, "AU12": 1.5}', 50, 424], # Slight smile | |
| ['{"AU4": 1.5, "AU15": 1.5}', 50, 424], # Slight frown | |
| ], | |
| inputs=[au_params_input, steps_slider, seed_input], | |
| ) | |
| transform_btn.click( | |
| fn=transform_emotion, | |
| inputs=[input_image, au_params_input, steps_slider, seed_input], | |
| outputs=[output_image, status_text] | |
| ) | |
| with gr.Tab("📡 API Documentation"): | |
| gr.Markdown(""" | |
| ## 🔗 API Usage | |
| ### Gradio API Endpoint | |
| **POST** `https://gauravvjhaa-affecto-inference.hf.space/api/predict` | |
| ```python | |
| import requests | |
| import base64 | |
| # Prepare image | |
| with open("face.jpg", "rb") as f: | |
| image_base64 = base64.b64encode(f.read()).decode() | |
| # Call API | |
| response = requests.post( | |
| "https://gauravvjhaa-affecto-inference.hf.space/api/predict", | |
| json={ | |
| "data": [ | |
| f"data:image/jpeg;base64,{image_base64}", | |
| '{"AU6": 2.0, "AU12": 2.0}', | |
| 50, # steps | |
| 424 # seed | |
| ] | |
| } | |
| ) | |
| result = response.json() | |
| # Result image is in result["data"][0] | |
| ``` | |
| ### 🎭 Available Action Units (AU): | |
| **😊 Happiness:** | |
| - **AU6** (Cheek Raiser): 0-4 → Raises cheeks, squints eyes | |
| - **AU12** (Lip Corner Puller): 0-4 → Pulls lip corners up (smile) | |
| **😢 Sadness:** | |
| - **AU1** (Inner Brow Raiser): 0-4 → Raises inner brows | |
| - **AU4** (Brow Lowerer): 0-4 → Lowers brows (frown) | |
| - **AU15** (Lip Corner Depressor): 0-4 → Pulls lip corners down | |
| **😠 Anger:** | |
| - **AU4** (Brow Lowerer): 0-4 | |
| - **AU5** (Upper Lid Raiser): 0-4 | |
| - **AU7** (Lid Tightener): 0-4 | |
| - **AU23** (Lip Tightener): 0-4 | |
| **😮 Surprise:** | |
| - **AU1** (Inner Brow Raiser): 0-4 | |
| - **AU2** (Outer Brow Raiser): 0-4 | |
| - **AU5** (Upper Lid Raiser): 0-4 | |
| - **AU26** (Jaw Drop): 0-4 | |
| **😨 Fear:** | |
| - **AU1** (Inner Brow Raiser): 0-4 | |
| - **AU2** (Outer Brow Raiser): 0-4 | |
| - **AU4** (Brow Lowerer): 0-4 | |
| - **AU5** (Upper Lid Raiser): 0-4 | |
| - **AU20** (Lip Stretcher): 0-4 | |
| - **AU25** (Lips Part): 0-4 | |
| ### 📊 Example Combinations: | |
| ```json | |
| # Happy (smile) | |
| {"AU6": 2.5, "AU12": 3.0} | |
| # Sad | |
| {"AU1": 2.0, "AU4": 2.5, "AU15": 2.5} | |
| # Angry | |
| {"AU4": 3.0, "AU5": 2.0, "AU7": 2.5, "AU23": 2.0} | |
| # Surprised | |
| {"AU1": 3.0, "AU2": 2.5, "AU5": 3.0, "AU26": 2.5} | |
| # Fear | |
| {"AU1": 2.5, "AU2": 2.5, "AU4": 2.0, "AU5": 3.0, "AU20": 2.0, "AU25": 2.0} | |
| ``` | |
| ### ⚙️ Pipeline Details: | |
| 1. **Face Detection** → InsightFace detects face | |
| 2. **Cropping** → Crop to 512×512 with alignment | |
| 3. **Background Extraction** → BiSeNet segments face from background | |
| 4. **Diffusion** → MagicFace transforms expression | |
| 5. **Output** → 512×512 transformed face | |
| ### ⏱️ Performance: | |
| - **CPU**: ~10-15 minutes per image | |
| - **GPU (T4)**: ~15-20 seconds per image | |
| ### 📝 Notes: | |
| - Input images should contain at least one face | |
| - If multiple faces, largest face is used | |
| - AU values typically range 0-4 (higher = stronger) | |
| - Seed ensures reproducibility | |
| """) | |
| with gr.Tab("ℹ️ About"): | |
| gr.Markdown(""" | |
| ## 🎭 Affecto - MagicFace Implementation | |
| ### Architecture: | |
| This is a complete implementation of the **MagicFace** diffusion model for facial emotion transformation. | |
| **Components:** | |
| - **InsightFace**: Face detection and landmark extraction | |
| - **BiSeNet**: Face segmentation for background extraction | |
| - **Stable Diffusion 1.5**: Base diffusion model | |
| - **Custom UNets**: ID encoder + Denoising UNet from MagicFace | |
| - **Action Units**: FACS-based emotion parameters | |
| ### References: | |
| - **MagicFace Paper**: [arxiv.org/abs/2408.00623](https://arxiv.org/abs/2408.00623) | |
| - **MagicFace Code**: [github.com/weimengting/MagicFace](https://github.com/weimengting/MagicFace) | |
| - **HuggingFace Model**: [huggingface.co/mengtingwei/magicface](https://huggingface.co/mengtingwei/magicface) | |
| ### Project Info: | |
| - **Project**: Affecto - Facial Emotion Transformation System | |
| - **Implementation**: Complete MagicFace pipeline with preprocessing | |
| - **Hardware**: CPU (free tier) - upgrade to GPU for faster inference | |
| --- | |
| **Made by Gaurav Jha and Raj Shakya** 🎓 | |
| Built using Gradio, PyTorch, and Diffusers | |
| """) | |
| print("✅ Affecto - Real MagicFace API Ready!") | |
| print(f"🌐 URL: https://gauravvjhaa-affecto-inference.hf.space/") | |
| if __name__ == "__main__": | |
| demo.queue(max_size=5) # Queue for long-running tasks | |
| demo.launch(server_name="0.0.0.0", server_port=7860) | |