Explosive Weapons and Arms Detection with sIngular Classification (WARDIC) on Novel Weapon Dataset using Deep Learning: Enhanced OODA Loop

Anant Bhatt#,Email

Amit Ganatra#,Email

Devang Patel Institute of Advance Technology and Research, Charotar University of Science and Technology, Nadiad Petlad Road, Changa, 388421, Gujarat, India

*Corresponding Author

#These authors contributed to this work equally.

Abstract

Rising armed conflicts and violent agitations in populated areas have upstretched stark trepidations worldwide. Especially, the armed conflicts are increasingly being fought using explosive weapons, rifles, and lethal arms, endangering civilian lives and infrastructure. Concerns raised by the state actors accentuate the importance of identifying the miscreants in possession of explosive weapons to limit unlawful activities. Armed forces, law enforcement agencies, while carrying out operations in populated areas, demand the high availability of meaningful information promptly to shorten the OODA loop for effective decision-making by the military commanders. Existing research is limited to classifying revolvers, pistols, and knives and has privations to detect explosive weapons and firearms due to severe void of relevant datasets, computationally optimal weapon detection methodology, which imposes severe restrictions in instrumenting an effective system. Hence, we introduce two customized, high-balanced Novel Operational Weapon Arms Datasets - named NOWAD post detailed Exploratory Data analysis. We propose a state-of-the-Art methodology to implement a novel weapons and arms detection singular classifier -named (WARDIC) to identify explosive weapons and arms. The WARDIC - an augmented architecture of the ConvNets and Singular Classifiers showed promising performance in detecting weapons in surveillance feeds. The evaluation metrics show promising performance of the tuned WARDIC classifier (fusion of DenseNet-121 with the Isolation Forest) over traditional baselines models with the perfect scores of 100%. Cross validations of the classifier employing 5-Fold and 10-Fold CV showed accuracy scores of 99.27% and 99.46%, respectively, with linear complexity. Our experiments propose the State-of-the-Art Classifier formulation, which shortens the OODA loop for effective decision-making by significant improvement in computational complexity to instrument a quick response system. The WARDIC classifier distinctively supersedes the performance of the existing classifiers with enhanced computational speed for veristic operational scenarios.

Explosive Weapons and Arms Detection with sIngular Classification (WARDIC) on Novel Weapon Dataset using Deep Learning: Enhanced OODA Loop