Stochastic-Class-Based-Hard-Example-Mining-for-Deep-Metric-Learning (SCHEM)

Implementation of Stochastic Class-based Hard Example Mining for Deep Metric Learning in Tensorflow 2.1.0. MobilenetV2 is used instead of InceptionV1.

Guide:

1. Load CUB dataset into root directory CUB_200_2011. Images should be located in CUB_200_2011/images/*
2. Run Create_CUB_Dataset.py to reshape the CUB images and save them all as .npy files
3. Run Train_Model.py with parameters to start training