🧵 WaxFashionStableDiffusion

A custom trained Stable Diffusion model for generating African Wax Print-inspired fashion patterns.This model is part of an ongoing research and creative exploration into increasing representation of African fashion in AI. This project leverages Stable Diffusion to generate vibrant African Wax Print textile patterns. Trained on a curated synthetic dataset of wax print images, the model captures key stylistic elements such as bold color contrasts, geometric motifs, and cultural symbolism. It enables designers, researchers, and creators to generate fashion patterns using simple text prompts.

📌 Model Overview

    Model Type: Fine-tuned Stable Diffusion
     Framework: diffusers, PyTorch
    Resolution: 512x512 images
    Use Case: Textile pattern design, creative fashion generation, cultural pattern synthesis

🎯 Use Cases

    Generate synthetic wax fabric designs
    Aid in textile prototyping and visual merchandising
    Explore culturally rich fashion patterns using AI

📁 Dataset

    This model was trained using the DreamBooth method on a synthetic dataset of African Wax Print patterns.
    Name: AfricanWaxPatterns_5KDataset
    Type: Image dataset
    Size: ~5,000 curated samples
    Dataset Link: paceailab/AfricanWaxPatterns_5KDataset

🛠️ How to Use

    🔧 In Python (Colab or local):

from diffusers import StableDiffusionPipeline
import torch

model_id = "paceailab/WaxFashionStableDiffusion"

pipe = StableDiffusionPipeline.from_pretrained(
    model_id,
    torch_dtype=torch.float16
).to("cuda")

prompt = "Beautiful African Wax Pattern with orange and black design"
image = pipe(prompt, num_inference_steps=50).images[0]

image.save("wax_pattern.png")

📸 Sample Outputs

    Prompt Output

“Beautiful African Wax Pattern with orange and black design”

“Bright African wax print with red and blue flowers”

“Traditional wax pattern with geometric black and gold”

🔍 Training Details

    Training Batch Size : 1
    Train Steps : 400
    Learning Rate : 5e-6
    Resolution : 512
    Training prompt : "African Wax Patterns"

🗂️ Files & Versioning

The repository includes essential components for loading and using the WaxFashionStableDiffusion model:
    tokenizer/: Contains files for converting text prompts into tokens that the model can understand and process.
    feature_extractor/: Responsible for preprocessing inputs, such as resizing images or normalizing data, before feeding them into the model.
    scheduler/: Manages the denoising process during inference, controlling how noise is added or removed to generate high-quality images.
    README.md – Contains project overview, usage instructions, and relevant details.
    model_index.json – Metadata file that defines how the model can be loaded using diffusers.
    v1-5-pruned.ckpt / .safetensors – Core model weights, available in both .ckpt (Checkpoint) and .safetensors (a safer, more efficient format) for model inference.
    v1-inference.yaml – Configuration file containing settings used during model inference.
    text_encoder/, unet/, vae/, safety_checker/ – Directories containing trained model components, like the text encoder, UNet, VAE, and safety checker used to run the model.
    .gitattributes – Manages Git LFS (Large File Storage) for efficiently handling large model files.

📚 Related Resources

    🔗 Stable Diffusion, base model

    🔗 Diffusers Library

💡 Citation

    If you use this model, please cite or credit:

@misc{waxfashion2025,
  title={WaxFashionStableDiffusion},
  author={Pace AI Lab},
  year={2025},
  howpublished={\url{https://huggingface.co/paceailab/WaxFashionStableDiffusion}}
}