Text-To-Image Generation Using Stable Diffusion

This repositiory aims to fine-tune latest stable diffusion model on CIFAR-10 and Oxford-102 (Flower) dataset. The details of the repository have been listed below:

stable_diffusion_cifar.ipynb notebook contains the code to fine-tune CIFAR-10 on stable diffusion. Please download CIFAR-10 dataset and create the format of dataset as instructed in the notebook.
stable_diffusion_flower.ipynb notebook contains the code to fine-tune CIFAR-10 on stable diffusion. Please download Oxford-10 (Flower) dataset and create the format of dataset as instructed in the notebook.
train_text_to_image.py contains the hugging face wrapper for fine-tuning stable diffusion.
fid_cifar10.ipynb calculates the FID between generated test set and ground truth test set of CIFAR-10.
fid_flower.ipynb calculates the FID between generated test set and ground truth test set of Oxford-102 (Flower).
gan_test_cifar.ipynb evaluates the classification performance of one of the state of the art models on the generated CIFAR-10 dataset.
resize.py contains the code for resizing images
split_* contain code for splitting images into classes for the use of imageFolder.

Please contact sv2128@nyu.edu for any issues/bugs.

Name		Name	Last commit message	Last commit date
Readme.md		Readme.md
Text_To_Image_Report.pdf		Text_To_Image_Report.pdf
fid_cifar10.ipynb		fid_cifar10.ipynb
fid_flower.ipynb		fid_flower.ipynb
gan_test_cifar.ipynb		gan_test_cifar.ipynb
resize.py		resize.py
split_cifar10.py		split_cifar10.py
split_flowers.py		split_flowers.py
stable_diffusion_cifar.ipynb		stable_diffusion_cifar.ipynb
stable_diffusion_flower.ipynb		stable_diffusion_flower.ipynb
train_text_to_image.py		train_text_to_image.py