Feature detector and descriptor evaluation in human action recognition

Ling Shao, Riccardo Mattivi

Research output: Contribution to conferencePaper

54 Citations (Scopus)

Abstract

In this paper, we evaluate and compare different feature detection and feature description methods for part-based approaches in human action recognition. Different methods have been proposed in the literature for both feature detection of space-time interest points and description of local video patches. It is however unclear which method performs better in the field of human action recognition. We compare, in the feature detection section, Dollar's method, Laptev's method, a bank of 3D-Gabor filters and a method based on Space-Time Differences of Gaussians. We also compare and evaluate different descriptors such as Gradient, HOG-HOF, 3D SIFT and an enhanced version of LBP-TOP. We show the combination of Dollar's detection method and the improved LBP-TOP descriptor to be computationally efficient and to reach the best recognition accuracy on the KTH database.
Original languageEnglish
DOIs
Publication statusPublished - Jul 2010
EventCIVR 2010 - ACM International Conference on Image and Video Retrieval - Xi'an, China
Duration: 1 Jul 2010 → …

Conference

ConferenceCIVR 2010 - ACM International Conference on Image and Video Retrieval
Period1/07/10 → …

Fingerprint Dive into the research topics of 'Feature detector and descriptor evaluation in human action recognition'. Together they form a unique fingerprint.

Cite this