Abstract | ||
---|---|---|
Recognition of human actions in videos has been an active research area in computer vision community due to its important theoretical research significance and extensive practical application value. As the most commonly used local feature descriptor of video, dense trajectories were shown superior performance for action recognition on a variety of datasets. However, the high computational complexity and huge storage space requirements of the algorithm limit its application scenarios. This paper optimizes the action recognition algorithm based on improved dense trajectories feature, we use Vector of Locally Aggregated Descriptors (VLAD) for feature encoding, which can greatly reduce the computational complexity and avoid expensive hard disk access, at the same time, can also effectively reduce the loss of feature information and improve the recognition accuracy. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1016/j.procs.2019.01.213 | Procedia Computer Science |
Keywords | Field | DocType |
Action Recognition,VLAD,Dense Trajectory,Video Analysis | Space requirements,Computer science,Action recognition,Algorithm,Local feature descriptor,Trajectory,Computational complexity theory,Encoding (memory) | Conference |
Volume | ISSN | Citations |
147 | 1877-0509 | 0 |
PageRank | References | Authors |
0.34 | 0 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Wenwan You | 1 | 1 | 1.06 |
Junqi Guo | 2 | 61 | 15.07 |
Ke Shan | 3 | 0 | 0.68 |
Yazhu Dai | 4 | 0 | 1.01 |