Accurate iris segmentation and recognition using an end-to-end unified framework based on MADNet and DSANet

•An end-to-end unified iris recognition framework without normalization is proposed for more accurate iris segmentation and recognition.•MADNet combines multiple attention and improved skip connection to make better distinctions between iris and non-iris.•DSANet introduced spatial attention mechanism into the dense connection, which made the iris features of different samples more discriminative.•Plenty ablation experiments were used to demonstrate the effectiveness of MADNet and DSANet structures.•The proposed method can improve the performance of low-quality iris images that have not be provided the corresponding GT image most obviously.