This is the official Tensorflow implementation for our CIKM'22 paper Deep Self-Adaptive Hashing for Image Retrieval.
- Install dependencies
conda create -n dsah python=3.6
source activate dsah
pip install -r requirements.txt
- Download the VGG pretrained weights from here.
- Download the pre-extracted features, RGB data for CIFAR-10, FLICKR25K and NUS-WIDE datasets.
- Create an initial similarity matrix based on
W_create.py
for each datasets.
- Run
train_cifar_gpu.sh
orrun_cifar_gpu.py
to train the hash model, which will save the hash code during training. - Run
eval.py
to evaluate the retrieval performance of saved hash code.
If you find our work helps, please cite our paper.
@inproceedings{lin2021deep,
title={Deep Self-Adaptive Hashing for Image Retrieval},
author={Lin, Qinghong and Chen, Xiaojun and Zhang, Qin and Tian, Shangxuan and Chen, Yudong},
booktitle={Proceedings of the 30th ACM International Conference on Information \& Knowledge Management},
pages={1028--1037},
year={2021}
}