Eigen-Evolution Dense Trajectory Descriptors
| Autor: | Vinh Tran, Yang Wang, Minh Hoai Nguyen |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
Computer science
business.industry Deep learning 05 social sciences Pooling ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Pattern recognition 010501 environmental sciences 01 natural sciences ComputingMethodologies_PATTERNRECOGNITION Feature (computer vision) Encoding (memory) 0502 economics and business Trajectory Action recognition Artificial intelligence 050207 economics business 0105 earth and related environmental sciences |
| Zdroj: | FG |
| DOI: | 10.1109/fg.2018.00076 |
| Popis: | Trajectory-pooled Deep-learning Descriptors have been the state-of-the-art feature descriptors for human action recognition in video on many datasets. This paper improves their performance by applying the proposed eigen-evolution pooling to each trajectory, encoding the temporal evolution of deep learning features computed along the trajectory. This leads to Eigen-Evolution Trajectory (EET) descriptors, a novel type of video descriptor that significantly outperforms Trajectory-pooled Deep-learning Descriptors. EET descriptors are defined based on dense trajectories, and they provide complimentary benefits to video descriptors that are not based on trajectories. Empirically, we observe that the combination of EET descriptors and VideoDarwin outperforms the state-of-the-art methods on the Hollywood2 dataset, and its performance on the UCF101 dataset is close to the state-of-the-art. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|