RAG‐Net: ResNet‐50 attention gate network for accurate iris segmentation

Iris segmentation is an important step in the process of iris recognition. images collected under non-cooperative conditions always contain various noise, which a challenge for segmentation. Most U-Net-based methods have made great achievements However, this architecture lacks focusing on target structures varying shapes, and robustness segmenting objects with significant shape variations. In paper, we propose RAG-Net: efficient method based deep learning. contrast to many previous convolutional neural network (CNN)-based methods, adopted attention gate (AG) mechanism ResNet-50 U-Net improve accuracy, AG module was included skip connection part RAG-Net further identify salient feature regions prune responses, preserve only activations relevant required information, used performance. Using model, environment can be realized. The proposed trained evaluated using CASIA.v4-distance, CASIA.v4-thousand, UBIRIS.v2, MICHE-I databases. From view results, one effective methods.