Every moment counts: Dense detailed labeling of actions in complex videos. ArXiv e-prints

DMCA

Every moment counts: Dense detailed labeling of actions in complex videos. ArXiv e-prints (2015)

Cached

Download Links

[ai.stanford.edu]

by Serena Yeung , Olga Russakovsky , Ning Jin , Mykhaylo Andriluka , Greg Mori , Li Fei-fei

Citations:

1 - 1 self

BibTeX

@MISC{Yeung15everymoment,
    author = {Serena Yeung and Olga Russakovsky and Ning Jin and Mykhaylo Andriluka and Greg Mori and Li Fei-fei},
    title = {Every moment counts: Dense detailed labeling of actions in complex videos. ArXiv e-prints},
    year = {2015}
}

OpenURL

Abstract

Every moment counts in action recognition. A compre-hensive understanding of human activity in video requires labeling every frame according to the actions occurring, placing multiple labels densely over a video sequence. To study this problem we extend the existing THUMOS dataset and introduce MultiTHUMOS, a new dataset of dense la-bels over unconstrained internet videos. Modeling multi-ple, dense labels benefits from temporal relations within and across classes. We define a novel variant of long short-term memory (LSTM) deep networks for modeling these tempo-ral relations via multiple input and output connections. We show that this model improves action labeling accuracy and further enables deeper understanding tasks ranging from structured retrieval to action prediction. 1.

DMCA