| Summary: | 碩士 === 國立交通大學 === 光電工程研究所 === 107 === Traditional computer vision technologies only can define objects with obvious features such as geometric patterns, squares, and circles in object recognition. However, it is difficult to define the difference between cats and dogs only through simple features. "which is cat-A in a group of cats." is more difficult for the traditional computer vision recognition algorithm. However, the development of deep learning algorithms is popular and the high performance of the computer hardware is cheap. Everyone can easily buy great computer and use deep learning algorithms. Deep learning algorithms can define any objects which we wanted. The goal of this paper is creating a human flow analysis system using face features. The system will contain face identity recognition, face expression recognition and eye eye-locking recognition. This paper will explain how to do face recognition in the past, and then describe how the deep learning algorithm to do face recognition. we designed an experiment which contained of the three parts: face recognition, face expression recognition, eye-locking system. Finally, we will propose the performance of the system in three part and one application to do everything in our experiment.
|
|---|
