Datasets
Standard Dataset
Motion Detection for Surveillance Applications
- Citation Author(s):
- Submitted by:
- Jabulani Mpofu
- Last updated:
- Mon, 10/30/2023 - 11:48
- DOI:
- 10.21227/8v4c-a464
- Data Format:
- License:
223 Views- Categories:
- Keywords:
Abstract
This dataset, presents the results of motion detection experiments conducted on five distinct datasets sourced from changedetection.net: bungalows, boats, highway, fall and pedestrians. The motion detection process was executed using two distinct algorithms: the original ViBe algorithm proposed by Barnich et al. (G-ViBe) and the CCTV-optimized ViBe algorithm known as α-ViBe.
Within the provided main zip file, users will find the detection results for each of these algorithms applied to each dataset, with all images stored in the PNG format. Additionally, the zip file includes accompanying .txt files, which present the confusion matrices for each result folder pertaining to every dataset.
This is provided to help gather statistics from the detection results:
1. Go to changedetection.net > Datasets, to download the 2012 Dataset.
2. Go to changedetection.net > results, to download the necessary code (MATLAB or Python) for processing the detection results against the grountruth data.
3A. On MATLAB:
(i) Use "Current Folder" to move into the folder with the changedetection.net 2012 dataset.
(ii) Move the provided detection result images into their respective folder (the images are already
in the desired .png format).
(iii) Call processFolder(datasetPath, binaryRootPath)
ex : processFolder('C:\dataset', 'C:\results')
3B. On Python:
(i)Download the dataset and unzip all the files in the same folder. ex :
"c:\dataset"
You should get this :
c:\dataset\baseline\highway\input\
c:\dataset\baseline\highway\groundtruth\
c:\dataset\baseline\...
c:\dataset\...
c:\dataset\thermal\...
*** Please do not change or put anything else in the "dataset" folder ***
On the same level as dataset\, there is a 'results' folder filled with the
proper folder hierarchy. These folders are all empty and will contain your
results.
(ii) If you already have an executable for your motion detection method and
would like to run it on every video, search for the 'TODO' in
'processFolder.py' and set the path to your executable.
The first two arguments of your executable must be
- path to video. ex : "C:\dataset\baseline\highway"
- path to save the binary images. ex : "C:\results\baseline\highway"
(ii) If on the other hand you need to code your method, than follow these
steps :
The code in the "c++ code" folder is an OpenCV*-based template that may
help you code your method.
Put your code in 'YourMethod.h' and use the methods defined in
'YourMethod.cpp'
* OpenCV is a free library of programming functions for real time computer
vision. You can download it at http://sourceforge.net/projects/opencvlibrary/
(iii)Call "processFolder.py datasetPath binaryRootPath"
ex : "python processFolder C:\dataset C:\results"
** The two motion detection algorithms are evaluated using 7 metrics, namely (1)Recall, (2)Specificity, (3)FPR, (4)FNR, (5)PBC, (6)Precision, (7)FMeasure.
