RECOD: Resource-Efficient Camouflaged Object Detection for UAV-Based Smart Cities Applications
Abbas Khan, Mustaqeem Khan, Wail Gueaieb, Abdulmotaleb El Saddik, Guilia De Masi, Fakhri Karray · 2023
This paper introduces a novel method for detecting camouflaged objects in smart cities, termed Resource-Efficient Camouflaged Object Detection (RECOD). Detecting camouflaged objects is crucial for the proper functioning of various applications in smart cities, including traffic management, security surveillance, environmental monitoring, and infrastructure inspection. The proposed RECOD method is evaluated on four benchmark Camouflaged Object Detection datasets and a specialized image set of real-world camouflaged objects, demonstrating its potential to improve the efficiency of diverse real-world applications within the smart city context. Significantly, our study reveals that the proposed method not only achieves impressive results but a real-time performance of 70 FPS, rendering it well-suited for various real-world applications and potentially deployable on unmanned aerial vehicles. These findings indicate the significant utility of the RECOD method in detecting camouflaged objects and the potential to enhance the individual and overall functionality of smart cities.