Celebrity face labeling and visualization for web videos

Abstract

This project aims to help users overcome the problem that the users may face when watching a video. We have developed an Android application that can automatically label faces in videos based on the video metadata and face similarity for the user to learn more about the person appearing on the screen. The application located the face by reading the metadata and try to record the face patterns. Therefore it can recognize the face from the video even if there are no metadata. Face tracking is implemented to adjust the face position during the video is played. An information window is designed to let the user receive the information about the people appearing on the screen by performing a simple tap action. The information is downloaded from the Internet to ensure that only the most updated information will be shown to the user. In addition, the user retrieves more information by clicking a button. The application includes face detection, face tracking, face recognition and video grouping by the person's name. The algorithm and the design implemented are based on Google Vision and the face detector from Android System. The SQLite database is also used. The product of this project targets on any Android devices with Android 4.4 or above installed.