ViFoSe_Videosurveillance_Dataset_2026

COLLECTION ITEM:Article dataPROJECT TITLE:2026_PiccininiEtAl_ViFoSeDESCRIPTION OF THE FILES IN THE COLLECTION:This dataset contains surveillance-style video sequences capturing people moving in front of a static camera under controlled indoor/outdoor conditions. The data are intended for research in computer vision and image processing, with a focus on foreground extraction, motion detection, and binary mask segmentation tasks. The dataset can support the development and benchmarking of algorithms for background subtraction, human detection, and video-based scene understanding, in line with the ViFoSe project objectives.MAIN CONTACT FOR THIS WORK:Prof. Filippo Piccinini, PhD, University of Bologna & IRST IRCCS Meldola, Italy. Email: [e-mail] KEY OF THE VIDEOS:In order to systematically classify surveillance video sequences, the following legend has been adopted:- S/D: to indicate respectively a static or dynamic camera;- d/s: to specify whether the background is static or dynamic;- O (Object): to indicate the reference subject;- 1/2/3: number of moving subjects present in the scene;Classification of case studies:- 01 Continuous movement with the subject fully visible;- 02 Interrupted movement with the subject fully visible;- 03 Continuous movement with the subject partially visible;- 04 Interrupted movement with the subject partially visible;- 05 One subject with continuous movement and fully visible; two subjects with continuous movement and partially visible;- 06 One subject with continuous movement and fully visible; two subjects with interrupted movement and partially visible;- 07 One subject with interrupted movement and fully visible; two subjects with continuous movement and partially visible;- 08 One subject with interrupted movement and fully visible; two subjects with interrupted movement and partially visible;KEYWORDS:Computer VisionImage ProcessingVideo AnalysisObject segmentationForeground DetectionARTICLE (when using these files, please, cite the following article):Filippo Piccinini, Michele Sarneri, Luca Rinaldi, Abdelrahman Abdelaziz Mohamed, et al. ViFoSe, an user-friendly open-source post-processing tool for foreground segmentation in time-lapse videos. 2026.ABSTRACT OF THE ARTICLE:Time-lapse videos are commonly acquired in microscopy and outdoor imaging to study the behaviour of objects of interest (OOIs) under different stimuli. However, in most cases, the relevant information pertains only to the OOIs, considered as the foreground (FG, e.g. moving cells), while the background (BG, e.g. an inorganic scaffold) is often irrelevant. Today, a wide range of algorithms to segment the FG is available, spanning from post-processing to real-time solutions, accommodating dynamic/static cameras and/or BG, and catering to both general and case-specific scenarios. However, no freely available tool allows users to test different automatic solutions while also providing manual corrections for missed detections or false inclusions.In this work, we propose Video Foreground Segmentation (ViFoSe), a user-friendly, open-source, post-processing tool for FG segmentation in time-lapse videos. ViFoSe includes several automatic algorithms, but its key feature is the set of options provided for easily correcting missed or incorrect detections.To validate the tool, videos under controlled conditions were acquired. Specifically, we began by considering common real-world scenarios with a static camera and static BK. Then, the analysis has been extended to all possible variations involving a dynamic camera and/or dynamic BK.The experiments conducted demonstrate ViFoSe's ability to correctly segment OOIs, even when they cross each other or are partially hidden behind BG elements, in cases with a static camera and static BK. However, poor results were obtained when analysing videos with a dynamic BG (e.g., moving cells in hydrogels with debris or outdoor videos recorded in windy conditions). The ViFoSe source code, standalone applications for Windows, macOS, and Linux, user documentation, video tutorial, and the extensive testbed dataset are publicly available at: https://github.com/UniBoDS4H/ViFoSe/.COPYRIGHT:* Copyright (c) 2026,* Filippo Piccinini,* University of Bologna, Italy.* All rights reserved.* This material is free; you can redistribute it and/or modify it under the terms of the CC BY 4.0.* This material is distributed WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.