上一条: Activity Image-to-Video Retrieval by Disentangling Appearance and Motion
下一条: Exploiting Motion Information from Unlabeled Videos for Image Action Recognition