Fingerprint Recognition using Siamese Network

The purpose of this project is to train a One-Short Learning Siasmese Network to make it recognise the fingerprints of a user. This project has two implementations one using Tensorflow and Keras, other using PyTorch.

File Structure:-

• fingerprint.ipynb
• images/
  • p1/
    • thumb_edited.jpg
  • p2/
    • thumb_edited.jpg
  • p3/
    • thumb_edited.jpg
• model/
  • model_test5.h5
• processed_images/
  • x1.npy
  • x2.npy
  • y.npy
• test_images/
  • thumb_edited.jpg
• README.md

Data:-

The data being fingerprints of a particular person is private to them. So, I have not included a dataset. But the folder structure defines how the data needs to recide for the working. The data/files in the folders are only placeholders.

The images folder should contain all the individual person's fingerprints separated in folders (p1/, p2/, p3/). Each folder must contain 1 fingerprint images of a separate person. An example of how the image should look like is present in the folders.

The model folder will store the model that will be trained.

The processed_images folder will store the processed images in the form of numpy arrays.

The test_images folder will contain the images used to test your model.

The file fingerprint.ipynb is the python notebook where the code resides which is to be executed.

Implementation:-

Make sure to setup your data in the particular format and folder structure. The notebook has comments indicating what each cell does. You can then execute the cells in fingerprint.ipynb to see how everything works.

The notebook has 3 part, the first part is the implementation using Tensforflow and Keras. The second part is the code to add a new user to your system. The third part is the model implemented using PyTorch.

Result:-