Multimedia Datasets for Anomaly Detection: A Review

Abstract

Audio or video anomaly datasets play a crucial role in automated surveillance. The range of applications expand from outlier object/ situation detection to the detection of lifethreatening events. This research area has been active for more than 1.5 decades, and consequently, more and more datasets dedicated to anomalous actions and object detection have been created. Making use of these public anomaly datasets enable researchers to compare various anomaly detection frameworks with the same input data. This survey aims to address the lack of a comprehensive comparison and analysis of public datasets for audio, video, and audio-visual based anomaly detection. It also assists in selecting the best dataset for bench-marking frameworks. Additionally, we discuss gaps in the existing dataset and future direction insights toward developing multimodal anomaly detection datasets.

A SURVY of video datasets for anomaly detection and human activity recognition

“…We are speaking of VIRAT [3] There are several important aspects to investigate when considering a comparison of existing video data sets such as: scene Density, Example anomalies/event, Resolution, Camera motion, Total No. of Frames, number name of Reference [9,21].…”

“…Web (panic-escape and mob fighting are recognized as anomalies), Canoe (a boat happening once in the scene is treated as an anomaly), UMN (few individuals acting for quick evacuation is viewed as abnormality here) [8], Subway entrance/exit are some instances of such datasets (movement in the wrong direction regarded as an anomaly). [9], fig (2) shows type of anomaly datasets Categorization based on scene-type. Fig (3).…”

An adaptive combination algorithm based on deep learning and genetic algorithm for anomalous events detection

“…This dataset contains 13 diverse categories of an anomaly such as abuse, accidents, and fights, in addition to the normal category. In total it includes 1,900 videos divided into 800 normal videos and 810 anomaly videos for the training, whereas the testing had 150 normal videos and 140 anomaly videos [10], [26]. This collection includes over 129 hours of films at a resolution of 320x240 and 13 million frames [10], [24], [26].…”

Anomaly detection in surveillance videos based on H265 and deep learning

“…There are 1900 surveillance movies in the collection, with an approximately equal amount of abnormal and normal videos. The training set contained 810 abnormal and 800 normal samples, while the testing set contains the remaining 140 abnormal and 150 normal movies [14,39]. This collection has over 129 hours of videos at a 320x240 resolution, 13 million frames, these videos are different in their length [14,37,39].…”

“…The training set contained 810 abnormal and 800 normal samples, while the testing set contains the remaining 140 abnormal and 150 normal movies [14,39]. This collection has over 129 hours of videos at a 320x240 resolution, 13 million frames, these videos are different in their length [14,37,39]. The anomalies in this dataset have a significant impact on public safety, hence we chose it because it contains various abnormal events type.…”