Enhancing object detection in x-ray baggage screening using Kolmogorov-Arnold networks
Duygu Selin Ak, Büşra Küçükateş Yalçın, Şükrücan Taylan Işıkoğlu · 2025
As X-ray baggage scanning devices become more prevalent at critical checkpoints, deep learning methods are increasingly important for detecting prohibited items. Human inspection is often inconsistent and subjective, highlighting the need for automated solutions. Among various deep learning techniques, Kolmogorov-Arnold Networks (KAN) have shown promising results in object detection tasks, outperforming other methods with higher accuracy. This article focuses on using the KAN model to identify specific objects such as liquids, mobile phones, tablets, laptops, weapons, and knives in luggage scanned by X-ray devices. Additionally, we will compare KAN’s performance to traditional detection methods, aiming to enhance the reliability and effectiveness of object detection in security-sensitive environments like airports.