Weakly-Shared Deep Transfer Networks for Heterogeneous-Domain Knowledge Propagation

By Xiangbo Shu, Guo-Jun Qi, Jinhui Tang, Jingdong Wang. 2015

212 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Wenwu Zhu
2. Peng Cui
3. Zhangyang Wang
4. Gabriella Csurka
5. Shiyu Chang
6. Daixin Wang
7. Yang Liu
8. Yan Kang

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Ming-Hsuan Yang
3. Zhangyang Wang
4. Dacheng Tao
5. Trevor Darrell
6. Chunhua Shen
7. Yun Raymond Fu
8. Chen Change Loy

Citations in the corpus

• Domain Adaptation for Visual Applications: A Comprehensive Survey (2017, background): “The weakly-shared Deep Transfer Networks for Heterogeneous-Domain Knowledge Propagation [206] learns a domain translator function from a multi-modal source data that can be used to predict class labels in the target even if only one of the modality is present.”
• Secure and Efficient Federated Transfer Learning (2019, methodology): “In order to generate labels for target domain the following translator function from [16] is used:”
• Deep Domain Generalization With Structured Low-Rank Constraint (2018, background): “The second related works are DSN [39] and WSDTN [47], which are both designed for domain adaptation.”
• Visual and Semantic Knowledge Transfer for Large Scale Semi-Supervised Object Detection (2016, background): “general categories of TL have been proposed in previous work: homogeneous TL [13], [39], [40] in a single domain but with different data distributions in training and testing sets, and heterogeneous TL [41], [42], [43] across different domains or modalities.”
• Deep Asymmetric Transfer Network for Unbalanced Domain Adaptation (2018, methodology): “Figure 3 shows that when the source domain accuracy increases, the target domain accuracy of DATN sup increases much faster than that of WSDTN and CDLS.”
• Studying Very Low Resolution Recognition Using Deep Networks (2016, background): “lows for certain flexibility of each channel for domain-specific features. Beyond VLRR, the “partially coupled” architecture could be potentially applied to many other cross-domain recognition problems [27]. We have verified that the partially coupled architecture could lead to additional gains compare”
• Transfer Neural Trees for Heterogeneous Domain Adaptation (2016, methodology): “[28] presented a deep neural network structure, while co-occurrence cross-domain data are required for training their networks.”