UROP Project Listing System
 
Project Title:
Emotional and Disease Detection based on Real-time Facial Landmark Detection
Supervisor:CHENG Kwang-Ting Tim
Co-Supervisor:-
Quota:2
Project Description:Facial landmarks are key facial points such as nose tips, mouth corners. etc. In our research group, we have developed a facial landmark detector on mobile platform which achieves state-of-the-art accuracy and real-time performance and our landmark detector can perform shape-preserving 3D landmark detection. In this project, we plan to apply our facial landmark detection solution for emotion and disease detection for senior citizens and for mentally handicapped persons. We are collaborating with local care centers for senior citizens and for mentally handicapped persons and our partners will provide us video data for the target analysis. Facial expression indicates emotion and can be used to detect certain mental and physical diseases. Automated observation and supervision may help early diagnosis. In this project, we plan to answer key questions including how to collect meaningful visual data for analysis, how to develop accurate models for expression detection of target subjects, how to recognize subtle expressions, how to observe tiny change among noisy temporal data, and how to find correlation between emotions and facial expressions, and furthermore, between certain diseases (such as depression, stroke, etc.) and facial expressions.
Course type: UROP1100  
Applicant's Roles:Help conduct literature survey, collect data, develop models, develop software, conductive experiments based on the collected data and data from collaborating partners, communicating with collaborating partners for data collection, conveying initial results, getting feedback and refine the tasks and models. 
Applicant's Learning Objectives:The applicant can learn a number of highly relevant computer vision techniques, and their application to a real-world problem which has significant societal impact.  
Complexity of the project:Moderate 
 



Copyright © 2016 HKUST. All rights reserved.