@inproceedings{62defa139a0840b9b67d77f3b72d7053,
title = "RGB-D object recognition using the knowledge transferred from relevant RGB images",
abstract = "The availability of depth images provides a new possibility to solve the challenging object recognition problem. However, when there is not enough labeled data, we cannot learn a discriminative classifier even using depth information. To solve this problem, we extend LCCRRD method by kernel trick. First, we construct two RGB classifiers with all labeled RGB images from source and target domain. The significant samples for both classifier are boosted and the non-significant ones are inhibited by exploiting the relationship between two domains. In this process, the knowledge of source RGB classifier can be transferred to target RGB classifier effectively. Then to improve the performance of RGB-D classifier by applying the knowledge from source domain, the predicted results of RGB-D classifier are made consistent to target RGB classifier. Furthermore all the parameters are optimized in a unified objective function. Experiments on four cross-domain dataset pairs shows that our approach is indeed effective and promising.",
keywords = "Depth images, RGB-D object recognition, Transfer learning",
author = "Depeng Gao and Rui Wu and Jiafeng Liu and Qingcheng Huang and Xianglong Tang and Peng Liu",
note = "Publisher Copyright: {\textcopyright} Springer International Publishing AG 2017.; 24th International Conference on Neural Information Processing, ICONIP 2017 ; Conference date: 14-11-2017 Through 18-11-2017",
year = "2017",
doi = "10.1007/978-3-319-70136-3\_68",
language = "英语",
isbn = "9783319701356",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "642--651",
editor = "Derong Liu and Shengli Xie and Dongbin Zhao and Yuanqing Li and El-Alfy, \{El-Sayed M.\}",
booktitle = "Neural Information Processing - 24th International Conference, ICONIP 2017, Proceedings",
address = "德国",
}