CSC2529-Project

course project for CSC2529 at the University of Toronto.

Overview

This project aims to study and improve the result of the paper Designing An Illumination-Aware Network for Deep Image Relighting. The original author provides the main structure of the codebase, and we mainly focus on modifying the architectures in the network/arch folder.

Usage

• We strongly suggested using Google Collab for this repo. The default environment of Google Collab is suitable for this repo
• If they are not available create the environment by conda env create -f environment.yml, but be warned that some conflicts might occur
• Create data folder in the root directory with subfolders train, validation, and test
• Download dataset from VIDIT dataset and put data into according folders based on option files in options
• Run python train.py -opt options/[option file] for training
• The trained network will be saved to experiment/[option name] folder
• Move the pre-trained network to pretrained folder based on option files in options
• Run python test.py -opt options/[option file] for testing