by Srikar Babu (EE21B138) and Akshay Govind (ME22B102)
This repository contains code for fine-tuning SDXL using DreamBooth and LoRA, along with inference code. It also contains code for running inception score and FID metrics on generated images. The fine-tuned model weights for Bajirao's images (bajirao_weights.zip) and Rajinikanth's images (rajini_weights.zip) are provided; these can be used along with the inference notebook to reproduce our results.
We also include some generated outputs along with the captions used (in a txt file inside the respective folders). Also, added the datasets used for fine-tuning the model.
Two instances of training were performed:
- Training on Bajirao Mastani movie.
- Training on the custom Rajinikanth dataset provided.
Report: Link