X-Net: A Deep Convolutional Neural Model For X-Ray Threat Detection

Overview

This research proposes X-Net, a novel deep learning architecture that enhances airport security through the detection of threats in X-ray luggage scans. Scanning of luggage is a critical part of aviation safety but is alarmingly unreliable due to human error, endangering the safety of millions of airline passengers. To address this error, this study introduces several deep learning innovations engineered for baggage scan analysis and packages them into one model, called X-Net. X-Net employs a network of deep convolutional lateral stacks that combine vertical residual transpose blocks with inter-layer connections. This combination allows for multi-directional gradient flow, resulting in richer and more robust internal feature representations. These innovations enable X-Net to perform exceptionally well on real-world baggage scans, significantly enhancing public safety and potentially saving thousands of lives: X-Net detects luggage threats 400% more accurately and 91 times faster than a TSA officer. Moreover, this proposed approach provides novel and empirically useful deep learning tools that can strengthen other fields of computer vision, such as autonomous driving, medical imaging analysis, and biometric security.

Awards (1)

• 1st Place Medicine & Health/Behavioral Sciences

Competition history

• JSHS 2020 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.