Deep supervised hashing network with integrated regularisation. Issue 12 (23rd August 2019)

• Record Type: Journal Article
• Authors: Liao, Jianxin; Li, Baoran; Yang, Di; Wang, Jingyu; Qi, Qi; Wang, Jing
• Is Part Of: IET image processing. Volume 13:Issue 12(2019)
• Page Start: 2143
• Page End: 2151
• Publication Date: 2019-08-23
• Subjects: feature extraction -- image classification -- learning (artificial intelligence) -- file organisation -- indexing -- image representation -- image retrieval -- information retrieval -- binary codes; integrated regularisation system -- DSHIR system -- existing end‐to‐end deep hashing systems -- state‐of‐the‐art systems -- deep supervised hashing network -- approximate nearest neighbour search -- large‐scale multimedia retrieval tasks -- storage -- retrieval efficiency -- hashing methods -- image retrieval -- deep structures -- image representation -- good hash codes -- binary hash code -- existing methods -- deep hashing system -- compact binary codes; Image processing -- Periodicals; 621.36705
• DOI: 10.1049/iet-ipr.2018.6644
• Languages: English
• ISSNs: 1751-9659

Abstract

Abstract : Hashing has been widely deployed to approximate nearest neighbour search for large‐scale multimedia retrieval tasks due to storage and retrieval efficiency. State‐of‐the‐art supervised hashing methods for image retrieval construct deep structures to simultaneously learn image representation and generate good hash codes, and the key step among them is simultaneously learned feature representation and binary hash code. Existing methods use similarity and regularity loss to train deep hashing systems, but these two functions usually work together but not cooperative, which may lead to inadequate performance of the whole system. In this study, a new method for training deep hashing system to learn compact binary codes is presented. The deep supervised hashing network with integrated regularisation (DSHIR) system develop the zero division restriction as a new part of the loss function, which settles the problem of cooperatively guiding the system generate similarity preserving binary codes. DSHIR system also modifies the similarity handling loss to better extract features from image data, which promotes the performance compared to existing end‐to‐end deep hashing systems. Experiments show that DSHIR yields about 10 per cent higher mean average precision on CIFAR‐10 dataset, and also promote on other evaluation indexes compared with state‐of‐the‐art systems.