Code for fine-tuning Stable Diffusion with custom image dataset
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.
-
create_json.py - creates a metadata json file for custom hugging face dataset using a prexisting csv file
-
text_to_image_script.sh - script to create text-to-image fine-tuned model (uses metadata)
-
dreambooth_model.ipynb - jupyter notebook to create dreambooth fine-tuned model (does not use metadata, only images)
Louis Chapo-Saunders