Entry–Exit Video Surveillance: A Benchmark Dataset

Abstract

Techniques to automate video surveillance around places where cameras are forbidden due to privacy concerns are yet under-addressed. This can be achieved by building conceptual models and algorithms to investigate the credibility of monitoring of events using the video frames captured by mounting the cameras so as to have the view of the entrances of such camera-forbidden areas. Evaluation of these models and algorithms require standard datasets. The proposal here is to introduce a new benchmark dataset—“EnEx dataset” as no traces specific to the problem were found in the literature. The dataset comprises of video frames captured in 5 different locations accounting 90 entry–exit event pairs based on 9 different sequences involving 36 participants. Ground statistics of the dataset is reported. This work ventures a new sub-domain for research in the area of automated video surveillance.