Physical Knowledge Driven Multi-scale Temporal Receptive Field Network for Compressed Video Action Recognition.
Lijun HeMiao ZhangSijin ZhangFan LiPublished in: UbiComp/ISWC Adjunct (2021)
Keyphrases
- action recognition
- knowledge driven
- compressed video
- receptive fields
- multiscale
- human actions
- natural images
- activity recognition
- spatio temporal
- human detection
- image representation
- space time
- computer vision
- spatial and temporal
- biologically inspired
- neural network
- bitstream
- video content
- video frames
- video streams
- image segmentation
- computational complexity
- low level
- video sequences
- denoising