N
Nenghai Yu
Researcher at University of Science and Technology of China
Publications - 89
Citations - 2046
Nenghai Yu is an academic researcher from University of Science and Technology of China. The author has contributed to research in topics: Feature (computer vision) & Feature extraction. The author has an hindex of 20, co-authored 89 publications receiving 1166 citations. Previous affiliations of Nenghai Yu include Chinese Academy of Sciences & City University of Hong Kong.
Papers
More filters
Proceedings ArticleDOI
Cross-Modality Person Re-Identification With Shared-Specific Feature Transfer
TL;DR: Wang et al. as mentioned in this paper proposed a cross-modality shared-specific feature transfer algorithm (termed cm-SSFT) to explore the potential of both the modality-shared information and the modal-specific characteristics to boost the reID performance.
Book ChapterDOI
Zoom-Net: Mining Deep Feature Interactions for Visual Relationship Recognition
TL;DR: This work presents two new pooling cells to encourage feature interactions, and sheds light on how one could resolve ambiguous and noisy object and predicate annotations by Intra-Hierarchical trees (IH-tree).
Book ChapterDOI
Dual supervised learning
TL;DR: Dual supervised learning as discussed by the authors proposes to train the models of two dual tasks simultaneously, and explicitly exploit the probabilistic correlation between them to regularize the training process, which can improve the practical performances of both tasks.
Proceedings ArticleDOI
Stereoscopic Neural Style Transfer
TL;DR: In this article, the disparity loss is incorporated into the widely adopted style loss function by enforcing the bidirectional disparity constraint in non-occluded regions, and a feed-forward network is proposed by jointly training a stylization subnetwork and a disparity sub-network.
Proceedings ArticleDOI
Semantics Disentangling for Text-To-Image Generation
TL;DR: Zhang et al. as mentioned in this paper proposed a Siamese mechanism in the discriminator to learn consistent high-level semantics, and a visual-semantic embedding strategy by semantic-conditioned batch normalization to find diverse lowlevel semantics.