Extracting faces from cctv footage using Multi Task Cascaded Convolutional Neural Networks and OpenCV

In crime activities, CCTV footage implies a great role but, in most cases, the CCTV footages are too blurry and not clear enough to identify the criminals. This is why it is much important to detect specific faces from low-resolution CCTV footage, where there can be low light, Multi-Task Cascaded Convolutional Neural Networks can be used to extract criminal faces in those situations.

Demo 1

Demo 2

How to run this project

Steps:

1. Intall OpenCV. For intalling OpenCV, use pip install opencv-python.
2. Install MTCNN. For installing MTCNN, use pip install mtcnn.
3. Replace the video.mp4 file with your desired video. Keep the video file name just as it is.
4. Now run the python script video.py.
5. It will pop up a processing named window for capturing the faces.
6. The faces will be saved as JPG files in the folder named Extracted Faces.

Source of the video in this project:

https://www.youtube.com/watch?v=b3C18k7vAP8

Background Story of this cctv footage:

The leaders of the Chhatra League at BUET tortured and eventually assassinated Abrar Fahad, a second-year electrical and electronic engineering student (12 February 1998 – 7 October 2019). Fahad was battered to death with a blunt object, according to an autopsy.

References:

1. BUET student abrar murder verdict: 20 students get death. (2022). Retrieved 17 September 2022, from https://www.thedailystar.net/news/bangladesh/crime-justice/news/buet-student-abrar-murder-verdict-20-students-get-death-2913011

2. N. Zhang, J. Luo and W. Gao, "Research on Face Detection Technology Based on MTCNN," 2020 International Conference on Computer Network, Electronic and Automation (ICCNEA), 2020, pp. 154-158, doi: 10.1109/ICCNEA50255.2020.00040.

3. I. Culjak, D. Abram, T. Pribanic, H. Dzapo and M. Cifrek, "A brief introduction to OpenCV," 2012 Proceedings of the 35th International Convention MIPRO, 2012, pp. 1725-1730.