Portrait Image Relighting

Brown University ~ Computer Vision ~ May 2020

In this project I implemented a deep learning research paper with a team of friends. The paper utilized convolutional neural networks to take the lighting from one portrait image and apply it to another. The primary challenge we faced was the size of the dataset. With full resolution images being used for training and testing, we did not have the computational power to implement the research paper as written, so instead we modified the task. We reduced the image sizes and optimized the network to produce similar results while being far less computationally expensive. Below is our final paper that outlines our process and results.