Skip to main navigation Skip to search Skip to main content

Eigen-evolution dense trajectory descriptors

  • Yang Wang
  • , Vinh Quang Tran
  • , Minh Hoai Nguyen
  • Stony Brook University

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

4 Scopus citations

Abstract

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.

Original languageEnglish
Title of host publicationProceedings - 13th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages473-479
Number of pages7
ISBN (Electronic)9781538623350
DOIs
StatePublished - Jun 5 2018
Event13th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2018 - Xi'an, China
Duration: May 15 2018May 19 2018

Publication series

NameProceedings - 13th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2018

Conference

Conference13th IEEE International Conference on Automatic Face and Gesture Recognition, FG 2018
Country/TerritoryChina
CityXi'an
Period05/15/1805/19/18

Keywords

  • Action Recognition
  • Dense Trajectories
  • Eigen Evolution Pooling
  • Trajectory Pooling

Fingerprint

Dive into the research topics of 'Eigen-evolution dense trajectory descriptors'. Together they form a unique fingerprint.

Cite this