Rethinking Online Action Detection in Untrimmed Videos

Rethinking Online Action Detection in Untrimmed Videos: A Novel Online Evaluation Protocol

Baptista Ríos, M. (University of Alcala)
Lopez-Sastre, Roberto J. (University of Alcala)
Caba Heilbron, Fabian (Adobe Research)
van Gemert, J.C. (TU Delft Pattern Recognition and Bioinformatics)
Acevedo-Rodriguez, F. Javier (University of Alcala)
Maldonado-Bascon, Saturnino (University of Alcala)

2019

The Online Action Detection (OAD) problem needs to be revisited. Unlike traditional offline action detection approaches, where the evaluation metrics are clear and well established, in the OAD setting we find very few works and no consensus on the evaluation protocols to be used. In this work we propose to rethink the OAD scenario, clearly defining the problem itself and the main characteristics that the models which are considered online must comply with. We also introduce a novel metric: the Instantaneous Accuracy ( $IA$ ). This new metric exhibits an online nature and solves most of the limitations of the previous metrics. We conduct a thorough experimental evaluation on 3 challenging datasets, where the performance of various baseline methods is compared to that of the state-of-the-art. Our results confirm the problems of the previous evaluation protocols, and suggest that an IA-based protocol is more adequate to the online scenario. The baselines models and a development kit with the novel evaluation protocol will be made publicly available.

Computer vision
deep learning
evaluation
instantaneous accuracy
online action detection

http://resolver.tudelft.nl/uuid:6cfcb10d-b0a0-4100-a336-c26ac10c2101

https://doi.org/10.1109/ACCESS.2019.2961789

2169-3536

IEEE Access, 8, 5139 - 5146

green

Institutional Repository

journal article

© 2019 M. Baptista Ríos, Roberto J. Lopez-Sastre, Fabian Caba Heilbron, J.C. van Gemert, F. Javier Acevedo-Rodriguez, Saturnino Maldonado-Bascon