MTRFN: Multiscale Temporal Receptive Field Network for Compressed Video Action Recognition at Edge Servers.
Lijun HeMiao ZhangSijin ZhangLiejun WangFan LiPublished in: IEEE Internet Things J. (2022)
Keyphrases
- action recognition
- compressed video
- receptive fields
- multiscale
- natural images
- human actions
- activity recognition
- image representation
- human detection
- biologically inspired
- computer vision
- video quality
- compressed domain
- video streams
- image segmentation
- body parts
- spatio temporal
- super resolution
- video content
- bitstream
- machine learning
- input image
- wavelet coefficients
- wavelet transform
- image features
- multiresolution
- hidden layer