Introduction

This is the source code of our TCYB 2018 paper "SCH-GAN: Semi-supervised Cross-modal Hashing by Generative Adversarial Network", Please cite the following paper if you use our code.

Jian Zhang, Yuxin Peng and Mingkuan Yuan, "SCH-GAN: Semi-supervised Cross-modal Hashing by Generative Adversarial Network", IEEE Transactions on Cybernetics (TCYB), 2018. [PDF]

Usage

For NUSWIDE dataset:

1. Pretrain the model by using the code under pretrain directory (pretrain-nus): python train.py
2. Train the model by using the code under SCHGAN-nus: python train.py
3. Generate hash codes for query and database samples by using the code under SCHGAN-nus: python test.py

Wikipedia and MIRFlickr datasets are similar to the NUSWIDE dataset.

For more information, please refer to our TCYB paper.

Welcome to our Laboratory Homepage for more information about our papers, source codes, and datasets.