Deep learning-based Image Style Transfer with Generative Adversarial Network

The project aims to develop the deep learning-based algorithm that translate the image style of specific object to the reference style. Firstly, the proposed research focuses on identifying the accurate region in image for style transfer, and then translating the image style in that region. Current techniques about image style transfer are struggling to focus on translating the desired objects while keeping the rest of regions in the image unchanged. The competitive advantage gained by the new technologies developed through this project will help Crater Labs to further grow and expand its business in the computer vision, thus creating new employment opportunities in Canada.

Faculty Supervisor:

Jun Chen

Student:

Kangdi Shi

Partner:

Crater Labs

Discipline:

Engineering - computer / electrical

Sector:

University:

McMaster University

Program: