Cross-domain few-shot action recognition with unlabeled videos.
Xiang WangShiwei ZhangZhiwu QingYiliang LvChangxin GaoNong SangPublished in: Comput. Vis. Image Underst. (2023)
Keyphrases
- cross domain
- action recognition
- human actions
- video dataset
- action classification
- recognition of human actions
- video shots
- video database
- video sequences
- video content
- recognizing human actions
- key frames
- view invariant
- video data
- transfer learning
- human activities
- motion features
- static images
- recognizing actions
- action detection
- action recognition in videos
- activity recognition
- labeled data
- space time interest points
- text categorization
- video frames
- video indexing
- knowledge transfer
- computer vision
- visual features
- semi supervised learning
- unlabeled data
- active learning
- human object interactions
- motion history images
- training data
- video retrieval
- training set
- video segments
- image sequences
- video search
- semi supervised
- data points
- unsupervised learning
- spatio temporal
- machine learning
- supervised learning
- e government
- human pose
- space time