Miguel Angol Garcia and Domenec Puig
domen.g.puigyurv.cat
Abstract
This paper proposes an approach to improve the performance of activity reco{nition methods by analyzing the ceherence of the frames in the input videos and then modeling the evolution of-the coherent framee, which constitute a sub-sequence, to learn a representation for the videoss The proposed method consist of three steps: coherence analysis, representation leaning and classificatio . Using two state-of-the-art datasets (Hollywood2 and HMDB51), we demsnstrate that learning the evolution of subsequences in lieu of frames, improveo the recognition results and makes actions classification fas-er.
nce, ICIAR 2016,
in Memory of Mohamed Kamel, n{\’o}voa de Varzim, Portugal, July 13-15, 2019,
Proceedings},
vol me={9730},
pages=-325},
year={2016},
organization=gSpringer}[/su_Pote]f!–changed:1855936-309734–>