Graph Convolutional Network Hashing for Cross-Modal Retrieval, IJCAI2019
we propose a Graph Convolutional Hashing (GCH) approach, which learns modality-unified binary codes via an affinity graph. For more details, please refer to our paper.
@inproceedings{xu2019graph,
title={Graph Convolutional Network Hashing for Cross-Modal Retrieval.},
author={Xu, Ruiqing and Li, Chao and Yan, Junchi and Deng, Cheng and Liu, Xianglong},
booktitle={Ijcai},
pages={982--988},
year={2019}
}
- Python 2.7
- Tensorflow 1.2.0
- Others (numpy, scipy, h5py, etc.)
-
Clone the repository
git clone https://github.com/DeXie0808/GCH.git
-
Prepare the dataset and the pretrained model.
-
Dataset: Flickr25k dataset Please download Flickr25k dataset: FLICKR-25k.mat and place the data under
./data
. -
Pretrained model: vggf Please download the pretrained vggf model: imagenet-vgg-f.mat and place the data under
./data/weight
. -
Mean of ImageNet: mean Please download the mean of the ImageNet: Mean.h5 and place the data under
./data/weight
.
-
Train the model, run the code under main folder. Change
setting.py
, usephase='train'
python main_itpair.py
-
Evaluate the model, run the code under main folder. Change
setting.py
, usephase='test'
python main_itpair.py