Spaces:
Sleeping
Sleeping
File size: 3,229 Bytes
05f137b 22610c3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 | ---
title: CycleGAN Face-Sketch Converter
emoji: 🎨
colorFrom: blue
colorTo: purple
sdk: streamlit
sdk_version: "1.28.0"
app_file: app.py
pinned: false
---
# 🎨 CycleGAN Face-Sketch Converter
Convert face photos to sketches and vice versa using CycleGAN.
## 🚀 Deployment on Hugging Face Spaces
### Step 1: Prepare Your Models
After training, run this cell in your notebook:
```python
# Run this in your training notebook
export_for_huggingface(checkpoint_epoch=100) # Replace 100 with your best epoch
```
This creates two files in `deployment_models/`:
- `photo_to_sketch.pth` (~44 MB)
- `sketch_to_photo.pth` (~44 MB)
### Step 2: Create Hugging Face Space
1. Go to [Hugging Face Spaces](https://huggingface.co/spaces)
2. Click **"Create new Space"**
3. Configure:
- **Space name**: `cyclegan-face-sketch`
- **License**: Choose appropriate license
- **Select SDK**: **Streamlit**
- **Space hardware**: CPU Basic (free) or GPU for faster inference
### Step 3: Upload Files
Upload these files to your Space:
```
your-space/
├── app.py # Streamlit application
├── requirements.txt # Python dependencies
├── photo_to_sketch.pth # Model file (Photo → Sketch)
├── sketch_to_photo.pth # Model file (Sketch → Photo)
└── README.md # This file (optional)
```
### Step 4: Your Space will Auto-Build
Hugging Face will automatically:
1. Install dependencies from `requirements.txt`
2. Run your Streamlit app
3. Provide a public URL
## 📦 File Sizes
- `photo_to_sketch.pth`: ~44 MB
- `sketch_to_photo.pth`: ~44 MB
- **Total**: ~88 MB (well within free tier limits)
## 🎯 Features
- ✅ Photo to Sketch conversion
- ✅ Sketch to Photo conversion
- ✅ Auto-detection of input type
- ✅ Camera input support
- ✅ Download results
- ✅ Responsive UI
## 💡 Usage Tips
### For Best Results:
- Use clear, front-facing photos
- Ensure good lighting
- Images are automatically resized to 256x256
### Hardware:
- **CPU Basic (Free)**: ~3-5 seconds per image
- **GPU T4 (Paid)**: ~0.5 seconds per image
## 🔧 Local Testing
Before deploying, test locally:
```bash
# Install dependencies
pip install -r requirements.txt
# Run Streamlit app
streamlit run app.py
```
Then open http://localhost:8501
## 📊 Model Details
- **Architecture**: CycleGAN with ResNet-based generators
- **Input Size**: 256x256 RGB images
- **Training**: Unpaired face-sketch dataset
- **Loss Functions**:
- Adversarial loss (LSGAN)
- Cycle consistency loss
- Identity loss
## 🐛 Troubleshooting
### Space doesn't start:
- Check all files are uploaded correctly
- Verify file names match exactly in `app.py`
- Check Space logs for errors
### Out of memory:
- Use CPU Basic hardware
- Models are optimized for CPU inference
### Slow inference:
- Upgrade to GPU hardware in Space settings
- Or use batch processing
## 📝 License
[Your License Here]
## 🙏 Acknowledgments
- CycleGAN paper: [Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks](https://arxiv.org/abs/1703.10593)
- Person Face Sketches dataset: [Kaggle](https://www.kaggle.com/datasets/almightyj/person-face-sketches) |