Login (DCU Staff Only)
Login (DCU Staff Only)

DORAS | DCU Research Repository

Explore open access research and scholarly works from DCU

Advanced Search

A fine grained quality assessment of video anomaly detection

Zhou, Jiang orcid logoORCID: 0000-0002-3067-8512, McGuinness, Kevin orcid logoORCID: 0000-0003-1336-6477, Antony, Joseph orcid logoORCID: 0000-0001-6493-7829 and O'Connor, Noel E. orcid logoORCID: 0000-0002-4033-9135 (2022) A fine grained quality assessment of video anomaly detection. In: International Conference on Content-based Multimedia Indexing, 14-16 Sept 2022, Graz, Austria. ISBN 978-1-4503-9720-9

Abstract
In this paper we propose a new approach to assess the performance of video anomaly detection algorithms. Inspired by the COCO metrics we propose a quartile based quality assessment of video anomaly detection to have a detailed breakdown of algorithm performance. The proposed assessment divides the detection into five categories based on the measurement quartiles of the position, scale and motion magnitude of anomalies. A weighted precision is introduced in the average precision calculation such that the frame-level average precision reported in categories can be compared to each other regardless of the baseline of the precision-recall curve in every category. We evaluated three video anomaly detection approaches, including supervised and unsupervised approaches, on five public datasets using the proposed approach. Our evaluation shows that the anomaly scale introduces performance difference in detection. For both supervised and unsupervised methods evaluated, the detection achieve higher average precision for the large anomalies in scale. Our assessment also shows that the supervised multiple instance learning method is robust to the motion magnitude differences in anomalies, while the unsupervised one-class neural network method performs better than the unsupervised autoencoder reconstruction method when the motion magnitudes are small. Our experiments, however, also show that the positions of the anomalies have impact on the performance of the multiple instance learning method and the one-class neural network method but the impact on the autoencoder-based approach is negligible.
Metadata
Item Type:Conference or Workshop Item (Paper)
Event Type:Conference
Refereed:Yes
Uncontrolled Keywords:video anomaly detection; neural networks; evaluation metrics
Subjects:Computer Science > Artificial intelligence
Computer Science > Information technology
DCU Faculties and Centres:DCU Faculties and Schools > Faculty of Engineering and Computing > School of Computing
DCU Faculties and Schools > Faculty of Engineering and Computing > School of Electronic Engineering
Research Initiatives and Centres > INSIGHT Centre for Data Analytics
Published in: CBMI '22: Proceedings of the 19th International Conference on Content-based Multimedia Indexing. . Association for Computing Machinery (ACM). ISBN 978-1-4503-9720-9
Publisher:Association for Computing Machinery (ACM)
Official URL:https://doi.org/10.1145/3549555.3549569
Copyright Information:© 2022 The Authors. Open Access (CC-BY 4.0)
Funders:Science Foundation Ireland (SFI) under Grant Number SFI/12/RC/2289_P2, co-funded by the European Regional Development Fund.
ID Code:27856
Deposited On:18 Oct 2022 15:43 by Jiang Zhou . Last Modified 18 Oct 2022 15:43
Documents

Full text available as:

[thumbnail of cbmi2022.pdf]
Preview
PDF - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
2MB
Downloads

Downloads

Downloads per month over past year

Archive Staff Only: edit this record