Street Scene: A new dataset and evaluation protocol for video anomaly detection
This work addresses the need for better datasets and evaluation protocols in video anomaly detection research, though it is incremental as it builds upon existing methods.
The authors tackled the problem of limited datasets and flawed evaluation in video anomaly detection by introducing the Street Scene dataset and two new evaluation criteria, showing their novel baseline algorithm significantly outperforms existing state-of-the-art methods on this dataset.
Progress in video anomaly detection research is currently slowed by small datasets that lack a wide variety of activities as well as flawed evaluation criteria. This paper aims to help move this research effort forward by introducing a large and varied new dataset called Street Scene, as well as two new evaluation criteria that provide a better estimate of how an algorithm will perform in practice. In addition to the new dataset and evaluation criteria, we present two variations of a novel baseline video anomaly detection algorithm and show they are much more accurate on Street Scene than two state-of-the-art algorithms from the literature.