Stable-Diffusion-Fine-tuning-with-Dreambooth

Code for fine-tuning Stable Diffusion with custom image dataset

Overview

This code is utilised alongside the Diffusers library to produce a fine-tuned stable diffusion model capable of producing life-like generative images based on the custom image dataset provided.

Project Structure

1. create_json.py - creates a metadata json file for custom hugging face dataset using a prexisting csv file

2. text_to_image_script.sh - script to create text-to-image fine-tuned model (uses metadata)

3. dreambooth_model.ipynb - jupyter notebook to create dreambooth fine-tuned model (does not use metadata, only images)

Author

Louis Chapo-Saunders