From Selective Deep Convolutional Features to Compact Binary Representations for Image Retrieval

Convolutional Neural Network (CNN) is a very powerful approach to extract discriminative local descriptors for effective image search. Recent work adopts fine-tuned strategies to further improve the discriminative power of the descriptors. Taking a different approach, in this paper, we propose a novel framework to achieve competitive retrieval performance. Firstly, we propose various masking schemes, namely SIFT-mask, SUMmask, and MAX-mask, to select a representative subset of local convolutional features and remove a large number of redundant features from a feature map. We demonstrate that proposed masking schemes are effectively to address the burstiness issue and improve retrieval accuracy. Secondly, we propose to employ recent embedding and aggregating methods to further enhance feature discriminability. Additionally, we include a hashing module to produce compact binary image representations which are effective for the retrieval. Extensive experiments on six image retrieval benchmarks demonstrate that our proposed framework achieves state-of-the-art retrieval accuracy.