Human ear-side detection based on YOLOv5 detector and deep neural networks
Huy Nguyen-Quoc, Vinh Truong Hoang · 2021 International Conference on Decision Aid Sciences and Application (DASA) · 2021
Personal identification is always one of the most crucial tasks in computer vision. It has a wide application in many different fields, such as information security, demographic data collection, and e-commerce. With the popularity of biometric factors in identification, especially the distinctive shape of the human ear is able to solve tasks that other patterns can not do in several special circumstances. Therefore, we propose a new ear detection system using YOLOv5 to improve the current ear identification systems and fill the lack of fast ear detectors. By applying YOLOv5, we can locate multiple tiny ears with a short inference time. Moreover, we add an ear-side classifier in the detection pipeline to support other ear-related advanced tasks in the future. Two databases, namely CelebAsia and EarThai, are built for evaluating the proposed approach. The experimental results show the promising performance of our techniques.