| --- |
| tags: |
| - gan |
| - stylegan2 |
| - image-generation |
| - fashion |
| - wax-pattern |
| license: mit |
| library_name: torch |
| --- |
| |
| # 🧵 WaxFashionStyleGAN |
| A **custom-trained StyleGAN2 model** for generating **African Wax Print-inspired fashion patterns**. This model explores the intersection of AI and African fashion, aiming to increase the representation of African textile design in digital spaces. Trained on a synthetic dataset of African Wax Print images, the model generates vibrant, culturally rich fashion patterns. It's designed for designers, researchers, and creators to generate unique fashion patterns. |
|
|
| ### 📌 Model Overview |
| **Model Type**: Fine-tuned StyleGAN2 |
| **Framework**: PyTorch, StyleGAN2-ADA |
| **Resolution**: 1024x1024 images |
| **Use Case**: Fashion design, textile prototyping, creative AI pattern generation |
|
|
| ### 🎯 Use Cases |
| Generate synthetic African Wax fabric designs |
| Assist in textile prototyping and visual merchandising |
| Enable designers to explore culturally diverse fashion patterns |
|
|
| ### 📁 Dataset |
| This model was trained on a synthetic dataset of African Wax Print patterns using the StyleGAN2 architecture. |
| **Name**: AfricanWaxPatterns_5KDataset |
| **Type**: Image dataset |
| **Size**: ~5,000 curated samples |
| **Dataset Link**: paceailab/AfricanWaxPatterns_5KDataset |
|
|
| ### 🛠️ How to Use |
| 🔧 In Python (Colab or local): |
|
|
| ```py |
| !git clone https://github.com/researchpace/waxfashion.git |
| from huggingface_hub import hf_hub_download |
| import torch |
| import legacy |
| import dnnlib |
| import sys |
| sys.path.append('/content/waxfashion/stylegan2-ada-pytorch') |
| |
| # Load model |
| model_path = hf_hub_download("paceailab/Waxfashion_StyleGAN", "selected_models/styleGAN2ada_Africanwax.pkl") |
| |
| # Load the pre-trained StyleGAN2 model |
| device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') |
| with dnnlib.util.open_url(model_path) as f: |
| G = legacy.load_network_pkl(f)['G_ema'].to(device) # Load the generator |
| |
| # Generate and display image |
| import numpy as np |
| import PIL.Image |
| |
| def generate_image(seed=42): |
| z = torch.from_numpy(np.random.RandomState(seed).randn(1, G.z_dim)).to(device) |
| img = G(z, None) # Generate image |
| img = (img.permute(0, 2, 3, 1) * 127.5 + 128).clamp(0, 255).to(torch.uint8)[0].cpu().numpy() |
| return PIL.Image.fromarray(img) |
| |
| image = generate_image(seed=100) |
| image.show() |
| ``` |
|
|
| <h2>🔗 GitHub Repository</h2> |
| <p>For more information, code, please visit the GitHub repository:</p> |
| <ul> |
| <li><a href="https://github.com/researchpace/waxfashion" target="_blank">WaxFashion GitHub Repository</a></li> |
| </ul> |
|
|
| ### 🔍 Training Details for Example Selected StyleGAN2 Model |
| Training was done on university **High Performance Computer**. Explored various parameters |
| <img src="https://cdn-uploads.huggingface.co/production/uploads/662301442e1aa351ad6248c7/KYJ-dO8MPxyHlMOwmbztW.png" width="400"/> |
| **Gpus** : 1 |
| **Kimg** : 5000 |
| **Cfg** : stylegan2 |
| **Gamma** : 10 |
| **Metrics** : fid50k_full, kid50k_full, pr50k3_full, ppl2_wend, is50k |
| **Learning Rate** : 0.002 |
| **Training Batch Size** : 32 |
| **Map depth** : 8 |
|
|
| ### 📸 Sample Outputs for the above trained model |
|
|
| Prompt Output |
| ## At 1000 kimg |
| <img src="https://cdn-uploads.huggingface.co/production/uploads/662301442e1aa351ad6248c7/OVJ2Z0mOfF8SQ8PWJl60V.png" width="300"/> |
|
|
| ## At 2000 kimg |
| <img src="https://cdn-uploads.huggingface.co/production/uploads/662301442e1aa351ad6248c7/5vqEZrYw46DZdb6cJ4Oxl.png" width="300"/> |
|
|
| ## At 3000 kimg |
| <img src="https://cdn-uploads.huggingface.co/production/uploads/662301442e1aa351ad6248c7/yEUPxKCFGMIGK6cmYmNvI.png" width="300"/> |
|
|
| ## At 4000 kimg |
| <img src="https://cdn-uploads.huggingface.co/production/uploads/662301442e1aa351ad6248c7/ZaUqp-S4FmJmMHdvfvhWe.png" width="300"/> |
|
|
| ## At 5000 kimg |
| <img src="https://cdn-uploads.huggingface.co/production/uploads/662301442e1aa351ad6248c7/kE8ZBc0GEsPl0mKZ6W26I.png" width="300"/> |
|
|
| <h2>📊 Evaluation Metrics</h2> |
| <p>The performance of the <strong>WaxFashionStyleGAN</strong> model is evaluated using the following metrics:</p> |
| <ul> |
| <li><strong>FID50k_full</strong>: Frechet Inception Distance (FID) calculated over 50,000 images. A decrease in FID indicates better quality and diversity of generated images.</li> |
| <li><strong>KID50k_full</strong>: Kernel Inception Distance (KID) calculated over 50,000 images. A decrease in KID indicates better quality of the generated images.</li> |
| <li><strong>PR50k3_full</strong>: Precision-Recall (PR) score for 50,000 images. An increase in PR score indicates better coverage of the image distribution.</li> |
| <li><strong>PPL2_wend</strong>: Perplexity score at the 2nd generation stage. A higher value indicates better diversity in the generated images.</li> |
| <li><strong>IS50k</strong>: Inception Score (IS) calculated over 50,000 images. An increase in IS indicates that the generated images are diverse and of high quality.</li> |
| </ul> |
|
|
| <h2>🗂️ Files & Versioning</h2> |
| <p>This repository includes the necessary files for using the <strong>WaxFashionStyleGAN</strong> model:</p> |
| <ul> |
| <li><strong>stylegan2-ada-pytorch/</strong>: Contains the core StyleGAN2 implementation, including: |
| <ul> |
| <li><strong>dnnlib/</strong>: Helper library for managing configurations and other utilities.</li> |
| <li><strong>torch_utils/</strong>: Contains PyTorch-specific utilities for model manipulation.</li> |
| <li><strong>training/</strong>: Scripts for model training.</li> |
| <li><strong>metrics/</strong>: Contains code for calculating evaluation metrics like FID.</li> |
| <li><strong>generate.py</strong>: Script for generating images from the trained model.</li> |
| <li><strong>train.py</strong>: Main script to start training the StyleGAN2 model.</li> |
| <li><strong>trainv1.py</strong>: Alternate training script for different configurations.</li> |
| <li><strong>legacy.py</strong>: Code for compatibility with older versions of StyleGAN models.</li> |
| <li><strong>style_mixing.py</strong>: Script for mixing styles between different models or generated images.</li> |
| <li><strong>projector.py</strong>: Tool to project images into the latent space of the StyleGAN2 model.</li> |
| </ul> |
| </li> |
| <li><strong>selected_models/</strong>: Directory for storing pre-trained model checkpoints.</li> |
| <li><strong>metrics_graph.ipynb</strong>: Jupyter notebook for analyzing training metrics and model performance.</li> |
| <li><strong>config.yml</strong>: Configuration file for training parameters.</li> |
| <li><strong>requirements.txt</strong>: Lists dependencies for running the model.</li> |
| <li><strong>environment.yml</strong>: Conda environment file to set up the development environment.</li> |
| <li><strong>.gitattributes</strong>: Git LFS management for large model files.</li> |
| <li><strong>model_index.json</strong>: Metadata file for model configuration and indexing.</li> |
| </ul> |
| |
|
|
| ### 📚 Related Resources |
| 🔗 [StyleGAN2 Paper](https://arxiv.org/abs/1912.04958) |
| |
| 🔗 [StyleGAN2-ADA GitHub](https://github.com/NVlabs/stylegan2-ada-pytorch) |
|
|
| ### 💡 Citation |
| If you use this model, please cite or credit: |
|
|
| ```py |
| @misc{stylegan2025, |
| title={WaxFashionStyleGAN}, |
| author={Pace AI Lab}, |
| year={2025}, |
| howpublished={\url{https://huggingface.co/paceailab/WaxFashionStyleGAN}} |
| } |
| ``` |
