This repo containts a notebook that presents a practical session for fine-tuning Stable Diffusion so it learns to generate yourself. It was created for the lab sessions of the Tools and Applications of Artificial Intelligence module of the IARFID master from Universitat Politècnica de València. It is based on the notebook by TheLastBen.
You can open the notebook by clicking on .
The goal of this lab session is to fine-tune Stable Diffusion using DreamBooth, to create a model which is able to generate yourself.
For this practical session we need a collection of photos of ourselves. It is recommended to use between 15 to 20 pictures with the following characteristics:
- 2 to 3 full-body pictures.
- 3 to 5 pictures in which only half of our body appears.
- 8 to 12 portrait pictures.
It is important to have variability in the images: different backgrounds, light conditions, clothes we are wearing, etc.
- Overview of Colaboratory.
- Guide to Markdown.
- Importing libraries and installing dependencies.
- Saving and loading notebooks in GitHub.
- Rombach, R., Blattmann, A., Lorenz, D., Esser, P., & Ommer, B. (2022). High-Resolution Image Synthesis with Latent Diffusion Models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10684–10695.
- Ruiz, N., Li, Y., Jampani, V., Pritch, Y., Rubinstein, M., & Aberman, K. (2022). DreamBooth: Fine Tuning Text-to-image Diffusion Models for Subject-Driven Generation. arXiv preprint arxiv:2208. 2242.