Efficient Visual Fault Detection for Freight Train Braking System via Heterogeneous Self Distillation in the Wild

TL;DR

This paper introduces a lightweight heterogeneous self-distillation framework for efficient visual fault detection in freight trains, achieving high accuracy and speed with low resource consumption suitable for real-world railway safety applications.

Contribution

The paper presents a novel heterogeneous self-distillation approach with a lightweight backbone and a new knowledge neck, improving detection accuracy and efficiency in resource-constrained environments.

Findings

Achieves over 37 frames per second in fault detection.

Maintains highest accuracy compared to traditional distillation methods.

Uses less memory and has the smallest model size among competitors.

Abstract

Efficient visual fault detection of freight trains is a critical part of ensuring the safe operation of railways under the restricted hardware environment. Although deep learning-based approaches have excelled in object detection, the efficiency of freight train fault detection is still insufficient to apply in real-world engineering. This paper proposes a heterogeneous self-distillation framework to ensure detection accuracy and speed while satisfying low resource requirements. The privileged information in the output feature knowledge can be transferred from the teacher to the student model through distillation to boost performance. We first adopt a lightweight backbone to extract features and generate a new heterogeneous knowledge neck. Such neck models positional information and long-range dependencies among channels through parallel encoding to optimize feature extraction capabilities. Then, we utilize the general distribution to obtain more credible and accurate bounding box estimates. Finally, we employ a novel loss function that makes the network easily concentrate on values near the label to improve learning efficiency. Experiments on four fault datasets reveal that our framework can achieve over 37 frames per second and maintain the highest accuracy in comparison with traditional distillation approaches. Moreover, compared to state-of-the-art methods, our framework demonstrates more competitive performance with lower memory usage and the smallest model size.