The Implementation of DRFNet Network for iris recognition. The Network is reference Toward More Accurate Iris Recognition Using Dilated Residual Features. If you have any questions, please feel free to communicate.
The DET_curve.png was drawn after training with improved DRFNet using test data. And the Mask Iris images were obtained by independent network branch. The test data set is CASIA-Iris-Lamp, and this set is not involved in training.
According to my continuous experiments on iris recognition tasks, I recommend two high-quality papers. You can have a look when you have time
- Mis-classifified Vector Guided Softmax Loss for Face Recognition
- SphereFace2_ Binary Classification is All You Need for Deep Face Recognition