Knowledge Commons of Institute of Automation,CAS
Open Set Domain Adaptation with Zero-shot Learning on Graph | |
Zhang XY(张昕悦)1,2; Yang X(杨旭)1; Liu ZY(刘智勇)1 | |
2023-02 | |
会议名称 | International Conference on Advances in Computer Vision, Image and Virtualization |
会议日期 | 2023-4 |
会议地点 | 苏州 |
摘要 | Open set domain adaptation focuses on transferring the information from a richly labeled domain called source domain to a scarcely labeled domain called target domain, while classifying the unseen target samples as one unknown class in an unsupervised way. Compared with the close set domain adaptation, where the source domain and the target domain share the same class space, the classification of the unknown class makes it easy to adapt to the real environment. Particularly, after the recognition of the unknown samples, the model can either ask for manually labeling or further develop the classification ability of the unknown classes based on pre-stored knowledge. Inspired by this idea, we propose a model for open set domain adaptation with zero-shot learning on the unknown classes in this paper. We utilize adversarial learning to align the two domains while rejecting the unknown classes. Then the knowledge graph is introduced to generate the classifiers for the unknown classes with the employment of the graph convolution network (GCN). Thus the classification ability of the source domain is transferred to the target domain, and the model can distinguish the unknown classes in detail with prior knowledge. We evaluate our model on digits datasets and the result shows superior performance. |
七大方向——子方向分类 | 目标检测、跟踪与识别 |
国重实验室规划方向分类 | 虚实融合与迁移学习 |
是否有论文关联数据集需要存交 | 否 |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/52209 |
专题 | 多模态人工智能系统全国重点实验室_机器人理论与应用 |
通讯作者 | Yang X(杨旭) |
作者单位 | 1.中国科学院自动化所 2.中国科学院大学 |
推荐引用方式 GB/T 7714 | Zhang XY,Yang X,Liu ZY. Open Set Domain Adaptation with Zero-shot Learning on Graph[C],2023. |
条目包含的文件 | 下载所有文件 | |||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
finalpaper-C211.pdf(791KB) | 会议论文 | 开放获取 | CC BY-NC-SA | 浏览 下载 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论