Unsupervised domain adaptation (UDA) on person Re-Identification (ReID) aims to transfer the knowledge from a labeled source domain to an unlabeled target domain. Recent works mainly optimize the ReID models with pseudo labels generated by unsupervised clustering on the target domain. However, the pseudo labels generated by the unsupervised clustering methods are often unreliable, due to the severe intra-person variations and complicated cluster structures in the practical application scenarios. In this work, to handle the complicated cluster structures, we propose a novel learnable Hierarchical Connectivity-Centered (HCC) clustering scheme by Graph Convolutional Networks (GCNs) to generate more reliable pseudo labels. Our HCC scheme learns the complicated cluster structure by hierarchically estimating the connectivity among samples from the vertex level to cluster level in a graph representation, and thereby progressively refines the pseudo labels. Additionally, to handle the intra-person variations in clustering, we propose a novel relation feature for HCC clustering, which exploits the identities from the source domain as references to represent target domain samples. Experiments demonstrate that our method is able to achieve state-of-the art performance on three challenging benchmarks.
Download full-text PDF |
Source |
---|---|
http://dx.doi.org/10.1109/TIP.2021.3094140 | DOI Listing |
J Stroke Cerebrovasc Dis
December 2024
Department of Neurology, the Second Affiliated Hospital, Harbin Medical University, Harbin, Heilongjiang 150000, China.. Electronic address:
Introduction: Branch atheromatous disease (BAD) is prone to early neurological deterioration (END), leading to a poor prognosis. The most common arteries causing END are the lenticulostriate arteries (LSA) and the paramedian pontine arteries (PPA). To gain insight into the characteristics of symptomatic plaques and their association with poor prognosis in patients with BAD, we conducted a prospective study using high-resolution magnetic resonance imaging (HRMRI).
View Article and Find Full Text PDFMed Image Anal
December 2024
School of Computer Science and Engineering, University of New South Wales, Sydney, NSW 2052, Australia.
Continual test-time adaptation (CTTA) aims to continuously adapt a source-trained model to a target domain with minimal performance loss while assuming no access to the source data. Typically, source models are trained with empirical risk minimization (ERM) and assumed to perform reasonably on the target domain to allow for further adaptation. However, ERM-trained models often fail to perform adequately on a severely drifted target domain, resulting in unsatisfactory adaptation results.
View Article and Find Full Text PDFQuant Imaging Med Surg
December 2024
Department of Radiology, Tianjin First Central Hospital, Tianjin, China.
Background: Dementia with Lewy bodies (DLB) and Alzheimer's disease (AD) are common forms of dementia, characterized by overlapping clinical symptoms. Functional neuroimaging can provide valuable information for precise diagnosis. Our objective was to explore cerebral perfusion alterations in DLB and AD, and to determine which perfusion parameters are helpful in distinguishing DLB and AD.
View Article and Find Full Text PDFBiophys J
December 2024
I.M. Sechenov Institute of Evolutionary Physiology and Biochemistry Russian Academy of Sciences, St. Petersburg, Russia; Department of Biochemistry and Biomedical Sciences, Master University, Hamilton, Canada. Electronic address:
Despite their large functional diversity and poor sequence similarity, tetrameric and pseudo-tetrameric potassium, sodium, calcium and cyclic-nucleotide gated channels, as well as two-pore channels, transient receptor potential channels and ionotropic glutamate receptors share a common folding pattern of the transmembrane (TM) helices in the pore-forming domain. In each subunit or repeat, the pore domain has two TM helices connected by a membrane-reentering P-loop. The P-loop includes a membrane-descending helix, P1, which is structurally the most conserved element of these channels, and residues that contribute to the selectivity-filter region at the constriction of the ion-permeating pathway.
View Article and Find Full Text PDFSensors (Basel)
December 2024
School of Integrated Technology, Yonsei University, Incheon 21983, Republic of Korea.
To improve the accuracy and reliability of LiDAR semantic segmentation, previous studies have introduced multi-modal approaches that utilize additional modalities, such as 2D RGB images, to provide complementary information. However, these methods increase the cost of data collection, sensor hardware requirements, power consumption, and computational complexity. We observed that multi-modal approaches improve the semantic alignment of 3D representations.
View Article and Find Full Text PDFEnter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!