Navigation
A team of IUB faculties and alumni have published an innovative paper on visual surveillance in a top-tier Q1 journal with Impact factor of 8.7, titled “Semi-supervised, Neural Network based approaches to face mask and anomaly detection in surveillance networks,” the paper appears in the esteemed Journal of Network and Computer Applications.
The team comprises IUB Computer Science and Engineering (CSE) alumni Anik Alvi, currently a graduate student at New Mexico State University (NMSU); Sabir Saheel, currently pursuing higher education at the University of Minnesota; and Aninda Roy; and Associate Professors Tarem Ahmed, PhD, and Faisal Uddin, PhD. This research work has been supported by the RIoT Research Center, Independent University, Bangladesh.
The research, driven by the complexities of surveillance in a world accustomed to mask-wearing during the pandemic, introduces an open-source autonomous surveillance system, marking a notable shift in the accessibility and efficiency of such technologies.
Alvi, Saheel and Roy began this project as their undergraduate thesis at IUB under the supervision of Dr. Ahmed and Dr. Uddin. The project was further developed during Alvi’s tenure as a graduate assistant at the New Mexico Water Resources Research Institute (WRRI) with contributions from his former classmates and supervisors.
The paper details the use of Multi-task Cascaded Convolutional Neural Networks (MTCNN) for facial feature detection, combined with a Gabor image feature extractor and a Kernel-based Online Anomaly Detection (KOAD) algorithm. This ensemble of technologies enables real-time identification of potential risks, enhancing security measures in various public and private settings. The research was rigorously tested across multiple datasets, including two from public online repositories. The results showed a remarkable 78% accuracy in mask detection, outperforming comparable algorithms.
Dr. Faisal Uddin expressed his gratitude towards IUB, especially the Vice Chancellor, for allowing the use of video data from the university's surveillance cameras, which played a crucial role in enhancing the real-world applicability of their research.
Q1 journal 1
Independent University, Bangladesh
