Supervised Online Hashing via Hadamard Codebook Learning

Wang

et al. 2022

Scientific Programming

When virtualizing large-scale images of the real world, online hashing provides an efficient scheme for fast retrieval and compact storage. It converts high-dimensional streaming data into compact binary hash codes while saving the structural characteristics between samples into the Hamming space. Existing works usually update the hashing function based on the similarity between input data, or design a codebook to assign code words for each single input sample. However, assigning code words to multiple samples while retaining the balanced similarity of the image instances is still challenging. To address this issue, we propose a novel discriminative similarity-balanced online hashing (DSBOH) framework in this work. In particular, we first obtain the Hadamard codebook that guides the generation of discriminative binary codes according to label information. Then, we maintain the correlation between the new data and the previously arrived data by the balanced similarity matrix, which is also generated by semantic information. Finally, we joined the Hadamard codebook and the balanced similarity matrix into a unified hashing function to simultaneously maintain discrimination and balanced similarity. The proposed method is optimized by an alternating optimization technique. Extensive experiments on the CIFAR-10, MNIST, and Places205 datasets demonstrate that our proposed DSBOH performs better than several state-of-the-art online hashing methods in terms of effectiveness and efficiency.

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Discriminative Similarity-Balanced Online Hashing for Supervised Image Retrieval

Wang

et al. 2022

Scientific Programming

“…We compare our method, OSSH, with the following online hashing methods: Online Sketching Hashing (OSH) [30], Online Kernel Hashing (OKH) [28], Adaptive Hashing (AdaptH) [27], Online Supervised Hashing (OSupH) [29], Mutual Information Hashing (MIH) [32] with the trigger update module, Hadamard Codebook based Online Hashing (HCOH) [24]. All the methods are provided by the authors.…”

Section: B Comparison In a Static Environmentmentioning

confidence: 99%

“…Inspired by the online learning methods, recently, online hashing methods [24]- [27] are proposed to learn the hash functions in the online settings by processing the data in a sequential order with one pass. According to the ways of learning the hash functions from the streaming data, the online hashing methods can be roughly categorized as the stochastic gradient decent (SGD)-based online hashing methods [27]- [29] and the data sketching-based online hashing methods [30], [31].…”

Section: Introductionmentioning

confidence: 99%

Online Supervised Sketching Hashing for Large-Scale Image Retrieval

Weng

Zhu

2019

IEEE Access

Online hashing methods have achieved a good tradeoff between the accuracy and the efficiency for learning the hash functions in the online settings. Compared to the stochastic gradient descent-based online hashing methods, the data sketching-based online hashing methods can preserve more information from the streaming data to learn the hash functions. However, the existing data sketching-based methods are unsupervised hashing methods which cannot utilize the supervised information to alleviate the gap between the binary codes and the semantic similarity. In this paper, we propose an online supervised hashing method in which a semantic similarity graph is designed to describe the data similarity according to the supervised information, and a data sketch is developed to preserve the characteristics of the graph with significantly smaller size in the online settings. Hence, the hash functions can be learned online from the data sketch with low computational time cost and small space storage. Furthermore, we develop a parallel computing version of the proposed method, aiming at accelerating the process of learning the hash functions. The experiments on three large-scale datasets show that our method can achieve comparable or better search accuracy compared to other online hashing methods.INDEX TERMS Online hashing, data sketching, supervised hashing.

“…To address the efficiency and effectiveness issues, hashing methods have become a hot research topic. A great number of hashing methods are proposed to map images into a hamming space, including traditional hashing methods (Andoni and Indyk 2006;Lin et al 2018;2019) and deep hashing methods (Cao et al 2017;2018b;Liu et al 2018;Sheng et al 2018). Compared with traditional ones, deep hashing methods usually achieve better re- Figure 1: To obtain the optimal boundary for points with similar hashing codes, we propose a novel self-paced deep adversarial hashing to generate hard samples, as shown in (b).…”

Section: Introductionmentioning

confidence: 99%

SSAH: Semi-Supervised Adversarial Deep Hashing with Self-Paced Hard Sample Generation

Sheng

Zhou

Liu

et al. 2020

AAAI

Deep hashing methods have been proved to be effective and efficient for large-scale Web media search. The success of these data-driven methods largely depends on collecting sufficient labeled data, which is usually a crucial limitation in practical cases. The current solutions to this issue utilize Generative Adversarial Network (GAN) to augment data in semi-supervised learning. However, existing GAN-based methods treat image generations and hashing learning as two isolated processes, leading to generation ineffectiveness. Besides, most works fail to exploit the semantic information in unlabeled data. In this paper, we propose a novel Semi-supervised Self-pace Adversarial Hashing method, named SSAH to solve the above problems in a unified framework. The SSAH method consists of an adversarial network (A-Net) and a hashing network (H-Net). To improve the quality of generative images, first, the A-Net learns hard samples with multi-scale occlusions and multi-angle rotated deformations which compete against the learning of accurate hashing codes. Second, we design a novel self-paced hard generation policy to gradually increase the hashing difficulty of generated samples. To make use of the semantic information in unlabeled ones, we propose a semi-supervised consistent loss. The experimental results show that our method can significantly improve state-of-the-art models on both the widely-used hashing datasets and fine-grained datasets.