Development of Deep Learning Models for Improvised Explosive Devices Detection

Improvised Explosive Devices (IEDs) are a significant danger to public safety, frequently resulting in critical damage and loss of lives when they explode without warning. Conventional methods for detecting these threats are often expensive and not very efficient. This study focuses on applying image processing techniques to detect explosive hazards by utilizing deep learning models such as Densenet121, MobilenetV2, and InceptionV3. These models are trained to spot explosives and separate them from non-explosive objects, offering a more efficient and effective solution compared to traditional detection methods. The research examines how effectively Densenet121, MobilenetV2, and InceptionV3 can detect IEDs, with a focus on their ability to tell apart explosive threats from safe objects. A dataset is utilized, which goes through initial processing, data boosting, and normalization to maximize the model’s performance. Densenet121 proves to be the most effective, achieving an accuracy of 97.37%, which is higher than that of MobilenetV2 and InceptionV3. Densenet121 features an advanced design and an efficient structure that allow it to achieve high accuracy while operating with low processing requirements. This blend of precision and efficiency makes it a strong and practical option for detection tasks. Densenet121 marks a significant advancement in automated security systems. Its ability to boost threat detection and promote public security makes it an essential tool for overcoming the challenges associated with explosive devices.