M2guda

Abstract

Cross-modal hashing is a critical but very challenging task that is to retrieve similar samples of one modality via queries of other modalities. To improve the unsupervised cross-modal hashing, domain adaptation techniques can be used to support unsupervised hashing learning by transferring semantic knowledge from labeled source domain to unlabeled target domain. However, there are two problems that cannot be ignored: (1) most of domain adaptation based researches mainly focused on unimodal hashing or cross-modal real value-based retrieval but the study for cross-modal hashing is limited; (2) most existing studies only consider one or two consistency constraints during the domain adaptation learning. To this end, this paper propose a novel end-to-end framework to realize unsupervised domain adaptation for cross-modal hashing. This method, dubbed M 2 GUDA, including four different consistency constraints: structure consistency, domain consistency, semantic consistency and modality consistency for domain adaptation learning. Besides, to enhance the structure consistency learning, we develop a multimetrics graph modeling method to capture structure information comprehensively. Extensive experiments are performed on three common used benchmarks to evaluate the effectivity of our method. The results show that our method outperforms several state-of-theart cross-modal hashing methods. CCS CONCEPTS• Information systems → Multimedia and multimodal retrieval.

DUSA: Decoupled Unsupervised Sim2Real Adaptation for Vehicle-to-Everything Collaborative Perception

“…Unsupervised Domain Adaptation (UDA) aims to produce a robust model that can generalize effectively to target domains with labeled source data and unlabeled target data. Many works [5,6,8,9,14,16,35,36,47,50,52,53] utilize adversarial learning [11] to align feature distributions across different domains by minimizing the H-divergence [2] or the Jensen-Shannon divergence [12] between two domains. Other kinds of methods [17,34,47,54] develop a variety of pseudo labels for unlabeled target domains to achieve self-training [20].…”

Minimally invasive trabecular meshwork surgery for open-angle glaucoma

“…Open-angle glaucoma (OAG) is the most common type, with an estimated global prevalence of 2.4% that equates to an a ected population of nearly 69 million people (Zhang 2021). In one large population cohort, one in six people with OAG became bilaterally blind (Peters 2013).…”

CREAMY: Cross-Modal Recipe Retrieval By Avoiding Matching Imperfectly

“…This work concentrates on a hot spot in food computing area, namely cross-modal recipe retrieval that aims to retrieve the corresponding food images by queries of recipes or vice versa. Unlike the simple image-text pair in traditional crossmodal retrieval [5]- [9], the samples in cross-modal recipe retrieval are much more complex. Specifically, the images are photos of cooked food according to the recipes consisting of three textual components: (1) title, a single sentence naming the food; (2) ingredients, a list of sentences to presents the needed ingredients for the food; (3) instructions, a list of sentences to describe the cooking steps in detail.…”