J
Jun Yu
Researcher at Hangzhou Dianzi University
Publications - 193
Citations - 10327
Jun Yu is an academic researcher from Hangzhou Dianzi University. The author has contributed to research in topics: Computer science & Deep learning. The author has an hindex of 38, co-authored 179 publications receiving 7667 citations. Previous affiliations of Jun Yu include Xiamen University & Jiangnan University.
Papers
More filters
Proceedings ArticleDOI
PCPCAD: Proposal Complementary Action Detector
TL;DR: This paper presents a novel proposal complementary action detector (PCAD) to deal with video streams under continuous, untrimmed conditions and learns an efficient classifier to classify the generated proposals into different activities and refine their temporal boundaries at the same time.
Posted Content
Improving White-box Robustness of Pre-processing Defenses via Joint Adversarial Training.
TL;DR: JATP as discussed by the authors proposed a joint adversarial training based pre-processing method to mitigate the robustness degradation effect of adversarial adversarial examples in a white-box setting.
Journal ArticleDOI
Semantic Disentanglement Adversarial Hashing for Cross-Modal Retrieval
TL;DR: Wang et al. as discussed by the authors proposed a deep cross-modal hashing method, called Semantic Disentanglement Adversarial Hashing (SDAH), which decouple the original features of each modality into modality-common features with semantic information and modality private features with disturbing information.
Posted Content
Comprehensive Graph-conditional Similarity Preserving Network for Unsupervised Cross-modal Hashing.
TL;DR: Wang et al. as mentioned in this paper proposed a deep graph-neighbor coherence preserving network (DGCPN) to address the inaccurate similarity problem by exploring and exploiting the data's intrinsic relationships in a graph.
Proceedings ArticleDOI
Triple Disentangling Network for Unsupervised Domain Adaptation
TL;DR: A Triple Disentangling Network, to disentangle these three types of information and then predict the target labels merely using semantic information, which can not only effectively alleviate the negative transfer of outliers through disentangling instance information, but also disentangled se-mantic information more thoroughly by exploring discriminative structure knowledge.