This is the source code of our TCSVT 2017 paper "SSDH: Semi-supervised Deep Hashing for Large Scale Image Retrieval", Please cite the following paper if you use our code.

Jian Zhang and Yuxin Peng, "SSDH: Semi-supervised Deep Hashing for Large Scale Image Retrieval", IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), doi: 10.1109/TCSVT.2017.2771332, 2017. [PDF]

Our code is based on early version of Caffe, all the dependencies are the same as Caffe.

The proposed SSDH also uses the Pre-trained VGG_CNN_F model, which can be downloaded at Caffe model zoo, download this model and put it in example/semihash/Pre_trained folder.

Here we use MIRFlickr dataset for an example, under "data/flickr25k" folder, there are two list, you should resize MIRFlickr dataset according to those two list, so that Caffe can read the image data.

The flickr25k_semitrain_multi_label_h5.list is list of h5 file path. The h5 file contains multi label for the images, which is denoted as a vector, for example, (0,0,0,1,1,0,0,0,0,0,0,0,1,0 ...). For more details, see the code of the multi_label layer.

1. Edit Makefile.config to suit your system
2. Compile code: make all -j8
3. Training the model: example/semihash/train_1.sh. You may change train_1.sh to adjust the parameters such as GPU id and model saving location. This code will save the models in models/flickr25k
4. Generate hash codes for testing set: example/semihash/extratfea_flickr25k_12bit.sh. You can adjust script to change hash code saving location GPU id etc.

For more information, please refer to our TCSVT paper.

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