Machine-Learning-Based Artistic Photo Manipulation and Stylization on Mobile Devices

Recent advances in using machine learning for object recognition and image manipulation have resulted in a new and emerging market for mobile applications that use machine learning for creating a variety of new artistic expressions. This research will develop a framework for performing machine-learning-based photo and video manipulation on mobile devices with the goal of integrating it with the Generate Toolkit. This proposal follows previous MITACS internships between the same partners and further extends our objectives. In the previous iterations of this project, we developed a fast, on-device style transfer library for both iOS and Android systems and started developing a user-interface to train machine learning models for style transfer, facilitating the exploration of the parameter space of the models within a computer-assisted creativity paradigm. Our main objectives for this internship term involves three tasks. TO BE CONT'D

Intern: 
Omid Alemi
Faculty Supervisor: 
Philippe Pasquier
Province: 
British Columbia
Partner University: 
Program: