UniNet-Pytorch

Pytorch port of UniNet.

An accurate and generalizable deep learning framework for iris recognition.

Reference:
Zijing Zhao and Ajay Kumar, "Towards More Accurate Iris Recognition Using Deeply Learned Spatially Corresponding Features", Internation Conference on Computer Vision (ICCV), Spotlight, Venice, Italy, 2017.

• Paper
• Code

Install

• Python 3.6
• Pytorch 1.0+
• torchvision 0.2.2+
• opencv 3.4
• caffe(Optional)
• tqdm(Optional)

Code structure

• ICCV17_release
  • Source code and caffe model attached to the paper
• models
  • Source code and caffe model attached to the paper
• util
  • caffemodel2pth.py
    • Export the network parameters from caffemodel to pytorch pth format
  • normalize.py
    • Function of iris image normalization.
  • normalize_tool.py
    • Tool for iris normalization.
    • Left click to mark, right click to draw a circle (at least 3 points),'q' key to confirm, other keys to cancel
    • Iris first, pupil rear
  • segment.py
    • Iris image segmentation
• enroll_dataset.py
  • Register all images in the folder
• enroll_single.py
  • Register single image in the folder
• evaluation.py
  • Evaluation
• match.py
  • Match
• verify.py
  • Identify
  • Compare the extracted mat file with all mat files in the folder