Share on Facebook Tweet on Twitter Share on LinkedIn Share by email
How to Find Interesting Locations in Video: A Spatiotemporal Interest Point Detector Learned from Human Eye movements

W Kienzle, B Schölkopf, F Wichmann, and MO Franz

Abstract

Interest point detection in still images is a well-studied topic in computer vision. In the spatiotemporal domain, however, it is still unclear which features indicate useful interest points. In this paper we approach the problem by emphlearning a detector from examples: we record eye movements of human subjects watching video sequences and train a neural network to predict which locations are likely to become eye movement targets. We show that our detector outperforms current spatiotemporal interest point architectures on a standard classification dataset.

Details

Publication typeInproceedings
Published inPattern Recognition
URL10.1007/978-3-540-74936-3_41
Pages405-414
OrganizationMax-Planck-Gesellschaft
AddressBerlin, Germany
PublisherSpringer
> Publications > How to Find Interesting Locations in Video: A Spatiotemporal Interest Point Detector Learned from Human Eye movements