IcewineChen / mxnet-batch_hard_triplet_loss

mxnet version batch hard triplet loss

Mxnet-version batch hard triplet loss

Based on (https://arxiv.org/abs/1703.07737)

Based on some tricks from omoindrot's repository. (https://github.com/omoindrot/tensorflow-triplet-loss)

Introduction

• In this paper,authors propose a new format of triplet loss called batch hard. For more details of batch hard triplet loss, you can read (https://arxiv.org/abs/1703.07737).
• More efficient than triplet loss which proposed by facenet. More details in (https://arxiv.org/abs/1503.03832)
• Inplement the hard mining method and soft-margin
• Can be used in many tasks. Firstly I code this to do some research on re-id and image retrieval tasks.
• In the future maybe add batch all triplet loss. Compared to batch hard, sometimes it can make the experiment more efficient.

Architecture

1. Using resnetV2 to get 128-dimension embeddings
2. Using triplet loss to train embeddings
3. the network is defined in resnet.py

Requirements

The code has been tested with CUDA 8.0 and ubuntu 16.04.

• python3
• mxnet-cu80==1.3

how to train:
See parsers in train.py. Then Set your dataset path and some params of based resnet network.
The network has been defined in resnet.py.Batch_hard.py now has been deprecated.