Improving Hashing by Leveraging Multiple Layers of Deep Networks

Hashing methods usually consist of two crucial steps: encoding data with features and learning hash functions. Recently, some deep neural networks based hashing methods have been proposed and shown their efficiency as deep models can offer discriminative features. However, few deep hashing methods consider to leverage features from multiple layers. It is well known that different layers can provide different types of features, e.g., high, mid and low-level features, etc. Thus, a model is expected to obtain good performance if it could leverage the features from multiple layers simultaneously. Motivated by this, in this paper, we propose a novel technique to leverage different types of features from multiple layers of deep neural network, which can improve the accuracy. Experiments on real datasets show that the performance of end-to-end deep hashing is significantly enhanced; moreover, non-deep hashing can also benefit from our proposed technique of leveraging multiple layers' features.