MULTI-TASKING LEARNING FOR THE PROBLEM OF DETERMINING THE POSITION AND DIRECTION OF PERSON AT THE TIME

With many practical applications print human life today such as manufacturing surveillance cameras, analyzing and processing customer behavior, ..., the problem of face detection and head pose estimation on digital images is being noticed by many researchers. A large number of proposed deep learning models have state-of-the-art accuracies such as YOLO, SSD, and MTCNN, solving the problem of face detection or HopeNet, FSA-Net, and RankPmodelsodel used for head pose estimation problems.According to bigmany state-of-the-artmethods, the process of this task consists of 2 parts face detection to head pose estimation. These two steps are completely independent and do not share information with each other. This makes the model clear print setup but does not leverage most of the featured resources extracted in each model. In this thesis, we proposed the PoseMultitask model with the motivation to leverage the features extracted from the face detection model, sharing them with the head pose estimation branch to improve accuracy. Also, with the variety of data, the Euler angle domain representing the face is large, our model can predict results in the 360° Euler angle domain. Applying the multi-tasking learning method, the PoseMultitask model can simultaneously predict the position and direction of the human head. To increase the ability to predict the head direction of the model, we change the representation of the human face from the rom Euler angle to vectors from the Rotation matrix.