Triplet-net-keras

Implementation of Triplet Neural Network on keras

This code is inspired from adambielski's github repository https://github.com/adambielski/siamese-triplet, who implement the Triplet Neural Network on PyTorch, and also this Kaggle example https://www.kaggle.com/guichristmann/training-a-triplet-loss-model-on-mnist by Guilherme Christmann who use tensorflow to implement the triplet net.

Dependencies

Tensorflow 1.5

keras 2.2

sklearn 0.19.1

matplotlib 2.2.2

seaborn 0.9

Instruction

All the required codes are contained inside the jupyter-notebook

Summary

The data distribution of raw MNIST dataset

Before using the Triplet Neural Network, the raw date distribution looks like this:

The data distribution of MNIST dataset after applying Triplet Neural Newtork

After using the Tiplet Neural Network, the dataset forms different clusters according to their classes