A lightweight secondary iris localization algorithm with improved YOLOv7
The traditional iris location algorithm has many problems,such as poor location accuracy,sensitivity to noise interference,poor robustness,and slow location speed,which greatly limits the development of iris recognition.With the development of deep learning,the performance of iris location algorithm combined with convolution has been greatly improved,but there is still a lot of room for improvement and need for improvement.Based on YOLOv7 algorithm,this paper proposes a lightweight secondary iris location algorithm to improve the iris location requirements.Experiments were carried out on the JLU4.0 and CASIA-irisV4-Lamp data sets.When the IoU threshold was 0.9,the positioning accuracy reached 0.983 and 0.935,and the mAP was 95.41 and 89.07,respectively.Compared with the original framework,the indicators were improved by 4.14 and 3.18,respectively.At the same time,the model size was only 11.5%of the original network framework.The results show that the improved model has small size,excellent positioning speed,accurate positioning and high robustness.
万方数据
维普
