zsffq999 / DISH

Scalable Discrete Supervised Hash Learning with Asymmetric Matrix Factorization(DISH)

DISH

This repository is the implementation of Scalable Discrete Supervised Hash Learning with Asymmetric Matrix Factorization (Full paper version, the short paper version is to appear on ICDM'16).

Requirements

• numpy
• scipy
• caffe

Running the code

1. Download the cifar-10 data(origin image in RGB format and GIST descriptor) in npy format and extract them to the root folder (link)

• Run python dish-d.py (for deep hashing) or python dish-k.py (for kernel-based hashing).
• NOTE: you may run python resize_256.py and download the weights of VGG-16 net before using VGG net for training.