ECE285 Project B : Style Transfer

Description

This is project on Style Transfer is developed by team Paint.io composed of Ayon, Alakh, Nileesh and Avinash.

Code organization

• CycleGAN/demo.ipynb -- Run a demo of our CycleGAN code
• CycleGAN/train.ipynb -- Run the training of our CycleGAN model (as described in Section 4 of the report)
• CycleGAN/CycleGAN_experiments.ipynb -- Run experiments for CycleGAN (as described in Section 5)
• CycleGAN/models.py -- contain generator and discrimator classes
• CycleGAN/utils.py -- helper Classes and methods
• CycleGAN/my_nntools.py -- helper Classes needed for training CycleGAN
• CycleGAN/main.py -- script for training/testing CycleGAN
• Gatys_transfer/Demo.ipynb -- Run a demo of our Gatys style transfer code
• Gatys_transfer/GatysStyleTransfer.ipynb -- Run the training of our Gatys style transfer model with VGG (as described in Section 4)
• Gatys_transfer/GatysExperiment1.ipynb -- Run experiments for gatys style transfer with VGG (as described in Section 5)
• Gatys_transfer/AlexNetStyleTransfer.ipynb -- Run experiment for gatys style transfer with Alexnet (as described in Section 5)
• Gatys_transfer/images -- images used for gatys style transfer

Datasets

The datasets can be downloaded from https://people.eecs.berkeley.edu/~taesung_park/CycleGAN/datasets

Models

The pretrained models for CycleGAN can be found at : https://drive.google.com/drive/folders/1hbsJI9wa1f4-3xF1C5hwBdqewvD0KYsl?usp=sharing

Results

Cycle GAN

Gatys Style transfer

Acknowledgements

We thank Prof. Charles Deledalle and Inderjot Saggu for their guidance and support.