Darwintrees for Action Recognition

Albert Clapes, Tinne Tuytelaars, Sergio Escalera; Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 3169-3178

Abstract


We propose a novel mid-level representation for action recognition on RGB videos. We model the evolution of improved dense trajectory features not only for the entire video sequence, but also on subparts of the video. Subparts are obtained using a spectral divisive clustering that yields an unordered tree decomposing the entire cloud of trajectories of a sequence. We compute videodarwin on video subparts, exploiting more finegrained temporal information and reducing the sensitivity of the standard time varying mean of videodarwin. Next, we model the evolution of features through both frames of subparts and paths in tree branches. We refer to these mid-level representations as node-darwintree and branch-darwintree respectively. For classification, we construct a kernel for both mid-level and holistic videodarwin. Our approach achieves better performance than standard videodarwin and defines the current state-of-the-art on UCF-Sports and Highfive action recognition datasets.

Related Material


[pdf]
[bibtex]
@InProceedings{Clapes_2017_ICCV,
author = {Clapes, Albert and Tuytelaars, Tinne and Escalera, Sergio},
title = {Darwintrees for Action Recognition},
booktitle = {Proceedings of the IEEE International Conference on Computer Vision (ICCV) Workshops},
month = {Oct},
year = {2017}
}