3-Stream Convolutional Networks for Video Action Recognition with Hybrid Motion Field.
Wukui YangShan GaoWenran LiuXiangyang JiPublished in: MMSP (2018)
Keyphrases
- action recognition
- motion field
- human actions
- action classification
- video dataset
- spatial temporal
- action detection
- motion estimation
- motion vectors
- motion compensation
- optical flow
- recognizing human actions
- video sequences
- image sequences
- human activities
- recognition of human actions
- motion model
- computer vision
- space time interest points
- video data
- mid level
- activity recognition
- video streams
- vector field
- moving objects
- motion estimation algorithm
- image motion
- flow field
- video frames
- visual features
- video content
- spatio temporal
- atomic actions
- space time
- scene flow
- event detection
- stream processing
- human motion
- key frames
- high quality
- multimedia