 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
Fig: Sample Images of TUVD-CSA Dataset
|
|
|
The Tripura University Video Dataset for Crime Scene Analysis (TUVD-CSA) dataset is being created in the Computer Vision Laboratory of Department of Computer Science and Engineering, Tripura University (A Central University). The dataset provides a realistic diverse set of Indoor videos that were captured by Nikkon D5100 camera with 30 fps (Frames per second). Camera is positioned on a tripod stand of height 9-10ft from the ground level and at an angle of 600 to the camera tripod. We use Nikkor 18-55mm lens with Shutter speed 1/125 - 1/200 and Apertures of f/5.6 - f/8. The resolution of all the collected videos are 1920X1080 pixels. Till date the dataset contains 150 videos (more than 5,00,000 frames).
|
|
| Importance/Utilities of the Dataset: |
There are very less number of publically available for firearm detection. Datasets for firearm detection that can provide a balanced coverage in terms of number of firearms used, different complex background, mass firing scenario, sudden illumination change which is the most challenging task in case of indoor and outdoor condition, pan and tilt of firearms etc. is scanty. TUVD-CSA is a video dataset that includes those real time challenges. It is a video dataset that can be used to track the motion of firearm and the object carrying firearms. It has also implication in the area of area of video forensics which includes the analysis of the crime scene before the crime, during the crime and also after the crime happens. Captured videos are also divided into three categories, a) Before crime b) During crime c) After crime.
Application Area: Security and Surveillance, Video Forensic Analysis.
The key features of the designed dataset are as follows:
i. The primary feature of the dataset is, it's the only video dataset that contain moving object(s) with and without gun(s) in hand, collected in indoor & outdoor conditions.
ii. False positive images like object(s) carrying mobile, bottle, pen etc. are also collected to enrich non-gun class.
iii. Panning, tilting of guns with respect to the stationary objects are present.
iv. As the videos are captured from a certain distance and objects are moving horizontally and vertically hence, effect of scaling, rotation and occlusion are also present in the dataset.
v. Changing illumination is a major challenge which is also incorporated using the lights of rooms. Guns and objects both are affected by the illumination variation.
vi. Simple and complex backgrounds in indoor and outdoor conditions are considered in this dataset.
vii. The dataset have application in Forensics analysis. It incorporates frames of before, during and after crime is occurred.
viii. Videos of people with musk on their face are also considered in this dataset.
|
|
Dataset Statistic:
| Image Type |
Camera Model |
Video Format |
Total Videos |
Total Frames |
|
Indoor Condition |
Outdoor Condition |
Movie Clips |
|
|
|
| Visual |
Nikkon D5100 |
.MOV |
65 |
60 |
25 |
>50,00,000 |
|
|
|
|
| Dataset Releasing Year: |
Dataset still growing.
|
|
| Size of Dataset: |
20 GB
|
|
| Publications: |
Journals
[1] Rajib Debnath, Mrinal Kanti Bhowmik, "A Novel Framework for Automatic Localization of Gun Carrying by Moving Person Using Various Indoor Outdoor Mimic and Real time Views/Scenes" IET Image Processing, Volume: 14, Issue: 17, pp: 4663-4675, ISSN 1751-9667, Impact Factor: 1.995, Accepted on 5th January, 2021, DOI: 10.1049/iet-ipr.2020.0706.
Proceedings
[1] Rajib Debnath and Mrinal Kanti Bhowmik, "Deep Classification of Gun Carried by Moving Persons Using Proposed TUVD-CSA Dataset", Accepted in 8th International Doctoral Symposium on Applied Computation and Security Systems (ACSS), was held at the University of Calcutta, India, from 09th to 10th April, 2021, Published by Springer.
[2] Rajib Debnath and Mrinal Kanti Bhowmik, "Automatic Visual Gun Detection Carried by A Moving Person", Proceedings of 15th (IEEE) International Conference on Industrial and Information Systems (ICIIS), Indian Institute of Technology (IIT) Ropar, India, from 26-28 Nov, 2020, Published by IEEE. DOI: 10.1109/ICIIS51140.2020.9342681.
|
|
| Copyright: |
The TUVD-CSA dataset is the property of Tripura University. This data can only be used for research or academic purposes. Any commercial use of the data whatsoever or incorporation of the data into a larger database intended for public distribution must be done with the explicit written consent of the administrators.
|
|
| Created By: |
Research Group, Computer Vision Laboratory, Tripura University (A Central University).
|
|
|
For Download few Samples Click Here.
|
|
|
For Access/Download This TUVD-CSA full Dataset Click Here
|