Abstract
Damage detection remains a critical challenge, especially within the industrial automation sector, necessitating the development of advanced inspection technologies and their potential applications. Conventional industrial inspection methods are hindered by high costs and operational disruptions, motivating the development of innovative and efficient solutions. This paper introduces a novel, architecture-agnostic deep neural network (DNN) knowledge distillation (KD) method able to enhance vision-based damage detection performance even in challenging industrial environments. Our proposed method integrates foreground knowledge with feature KD to enhance data feature utilization in detection models, effectively minimizing background clutter. The results demonstrate the efficiency of our method in consistently enhancing the student’s training process, including up to a 12% increase in mean Aver-age Precision (mAP), across various DNN architectures. Our approach bridges the gap between academic research and real-world industrial cur-rent applications, offering a robust solution for damage detection in insulated pipelines.
Foreground-Aware Knowledge Distillation for Enhanced Damage DetectionThis work has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement number 101070604 – SIMAR.
The paper was presented by Pantelis Mentesidis at the 2nd workshop on Vision-based InduStrial InspectiON (VISION), 30/9/2024 at ECCV 2024, Milan, Italy
Publication not available yet.
Preprint available on the ZENODO SIMAR Community page