Properties to better characterize appropriate intermediate domains, we enforce Properties that these intermediate domains should satisfy. Based on the "shortest geodesic path"ĭefinition, i.e., the intermediate domains along the shortest geodesic pathīetween the two extreme domains can play a better bridging role, we propose two Representations using two domain factors. Representations on-the-fly by mixing the source and target domains' hidden Intermediate Domain Module (IDM) to generate intermediate domains' Intermediate domains to characterize this bridging. To tackle the UDA re-ID task, and we focus on explicitly modeling appropriate From a novel perspective, weĪrgue that the bridging between the source and target domains can be utilized Transferring the labeled source domain's knowledge to improve the model'sĭiscriminability on the unlabeled target domain. Authors: Yongxing Dai, Jun Liu, Yifan Sun, Zekun Tong, Chi Zhang, Ling-Yu Duan Download PDF Abstract: Unsupervised domain adaptive person re-identification (UDA re-ID) aims at
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |