This repository contains the code for the deep hashing approach proposed in our paper:
"Deep Supervised Hashing with Triplet Labels". Xiaofang Wang, Yi Shi and Kris M. Kitani. ACCV 2016.
Part of the code is modified from here.
We provide an example to run our method on the CIFAR-10 dataset. Our proposed deep hashing method can generate low-dimensional binary codes for images, which can later be used for large-scale image search. Details about the approach can be found in our paper "Deep Supervised Hashing with Triplet Labels".
This code is written in MATLAB and requires MatConvNet.
- Download the CIFAR-10 dataset from here. Uncompress the file and put the folder "cifar-10-batches-mat/" under the main folder.
- Download the Pretrained VGG-F model from here. Put the model under the main folder.
Run the following command in MATLAB:
$ DTSH(24)
This will gives you the binary codes generated by our method, as well as the test performance on the CIFAR-10 dataset. Here 24 represents the traget code length and you may change it to other numbers.
If you use the code, please cite:
@article{wang2016deep,
title={Deep Supervised Hashing with Triplet Labels},
author={Wang, Xiaofang and Shi, Yi and Kitani, Kris M},
journal={Asian Conference on Computer Vision},
year={2016}
}
- Xiaofang Wang (Robotics Institute, Carnegie Mellon University)
- xiaofan2@andrew.cmu.edu
- Personal Homepage