Whitehill, Jacob Richard
In recent times, the advancement in object detection has induced new video processing applications. This paper explores the development of the face tracking system with a probabilistic approach where face similarities and relative positions are considered. The potential applications and further enhancement includes detecting the human faces along with the behavior analysis and video surveillance at the chaotic situation such as objects overlapping. The system includes three stages - face detection and face alignment using MTCNN, face recognition using FaceNet, and face tracking after Gibbs sampling. I evaluate the full-temporal method and three other approaches - baseline, positional, and instantaneous methods.
Worcester Polytechnic Institute
Major Qualifying Project
All authors have granted to WPI a nonexclusive royalty-free license to distribute copies of the work, subject to other agreements. Copyright is held by the author or authors, with all rights reserved, unless otherwise noted.