A Highly Compressed Accelerator With Temporal Optical Flow Feature Fusion and Tensorized LSTM for Video Action Recognition on Terminal Device.
Peining ZhenXiaotao YanWei WangHao WeiHai-Bao ChenPublished in: IEEE Trans. Comput. Aided Des. Integr. Circuits Syst. (2023)
Keyphrases
- action recognition
- highly compressed
- feature fusion
- optical flow
- motion features
- human actions
- action classification
- computer vision
- motion history images
- bag of words
- activity recognition
- feature extraction
- multiple features
- human activities
- spatio temporal
- long video sequences
- image sequences
- space time
- atomic actions
- d scene
- motion field
- video frames
- feature set
- motion estimation
- viewpoint
- feature space
- image processing