Authentically Distorted Surveillance Videos Dataset

Pontificia Universidad Javeriana de CaliFaculty of Engineering and SciencesDepartment of Electronics and Computer ScienceElectronic Engineering Authors: César A. Ardila F. ([e-mail]) Santiago N. Bonilla V. ([e-mail]) Advisers: Hernán D. Benítez R. ([e-mail]) Roger A. Gomez N. ([e-mail]) ---------------------------------------------------------------------------------------------------------------------------- Abstract: This Dataset contains "Pristine" and "Distorted" videos recorded in different places. The distortions with which the videos were recorded are: "Focus", "Exposure" and "Focus + Exposure". Those three with low (1), medium (2) and high (3) levels, forming a total of 10 conditions (including Pristine videos). In addition, distorted videos were exported in three different qualities according to the H.264 compression format used in the DIGIFORT software, which were: High Quality (HQ, H.264 at 100%), Medium Quality (MQ, H.264 at 75%) and Low Quality (LQ, H.264 at 50%).The recording of the videos was carried out in a semicontrolled space, with the appropriate environment conditions that facilitate the generation of authentically distorted videos. The "DIGIFORT" monitoring software is used to manage and record the videos. For the labeling of each video we use the "DarkLabel" labeling software. For the videos editing and processing the "MATLAB" computational tool is used. The dataset contains 160 "Pristine" videos, 1450 "Exposure" videos, 1409 "Focus" videos and 1450 "Focus + Exposure" videos for a total of 4476 videos. The recorded activities are: "Walking" (WL), "Leaving Package in a Public Place" (LPP), "Passing Out" (PO), "Person Pushing Person" (PPP), "Person Running" (PR), "Fighting in Group" (FG), "Robbery with Knife" (RK) and "Prowl" (PW). This Dataset is intended to evaluate "Visual Quality Assessment" (VQA) and "Visual Object Tracking" (VOT) algorithms.