Breaking down violence detection: Combining divide-et-impera and coarse-to-fine strategies

Abstract

Breaking down violence detection: combining divide-et-impera and coarse-to-fine strategies. Neurocomputing, 208, pp. 225-237. (doi:10.1016Neurocomputing, 208, pp. 225-237. (doi:10. /j.neucom.2016 This is the author's final accepted version.There may be differences between this version and the published version. You are advised to consult the publisher's version if you wish to cite from it.http://eprints.gla.ac.uk/119160/

Learning to detect violent videos using convolutional long short-term memory

“…Several techniques have been proposed by researchers for addressing the problem of violence detection from videos. These include methods that use the visual content [21,2], audio content [23,12] or both [33,1]. In this section, we will be concentrating on methods that use the visual cues alone since it is more related to the proposed approach and moreover audio data is generally unavailable with surveillance videos.…”

Energy-Efficient Acoustic Violence Detector for Smart Cities

“…Such is the case of group G 8 (SF), where the difference between the appearance in both classifiers is always considerable. The opposite happens in groups G 3 and G 5 . In fact, the appearance of group G 3 in quadratic detector is always 0%.…”

“…This algorithm has been applied using a constraint related to the cost available in the system. Specifically, different cost thresholds measured in "Maximum number of Mega Floating Operations Per Second" (MaxMFLOPS) have been applied (1,3,5,10 and 15 MaxMFLOPS). This means that the sum of costs of the selected features has to be below this values.…”

“…For instance, when group G 1 -MFCCs and ΔMFCCs-appears (from 5 MaxMFLOPS onwards) its appearance is 100% in linear and quadratic detectors, while the appearance of the features that were selected previously is significantly reduced, like in groups G 4 and G 5 . Because of that, MFCCs is an excellent group.…”

“…In this sense, violence can be detected through audio and video surveillance. Some works in the literature treat this problem using both audio and video processing, 3,4,5 and the results obtained with the combination of those sources seems to be efficient.…”