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

Abstract

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.

Bridging Gap between Image Pixels and Semantics via Supervision: A Survey

“…It uses labeled data for empirical error minimization and both labeled and unlabeled data for embedding error minimization. The generative adversarial learning approach was also utilized in semi-supervised deep image retrieval [60,142]. A teacher-student semisupervised image retrieval method was presented in [172], where the pairwise information learned by the teacher network is used as the guidance to train the student network.…”

Clean-label Backdoor Attack against Deep Hashing based Retrieval

“…In labelinsufficient scenarios, deep hashing is designed for exploiting unlabeled or weakly labeled data, e.g. semi-supervised hashing (Yan, Zhang, and Li 2017;Jin et al 2020), unsupervised hashing (Shen et al 2018;Yang et al 2019), and weakly-supervised hashing (Li et al 2020;Gattupalli, Zhuo, and Li 2019). Moreover, building upon the merit of deep learning, hashing technique has also been applied in more challenging tasks, such as video retrieval (Gu, Ma, and Yang 2016) and cross-modal retrieval (Jiang and Li 2017).…”

Webly Supervised Deep Attentive Quantization

“…A recent milestone of learning to hash is the integration of hashing and deep learning. There have been many breakthroughs, both on the deep binary hashing [8,9] and deep quantization [10,11,12,13]. Despite the excitement to witness the harvest of deep hashing models that keep refreshing state of the art, we find that most of current models still relied on high-quality supervision.…”