Instance-Aware Knowledge Distillation for Semi-Supervised Learning of an On-Board Multi-Task Dense Prediction Model for Collision Avoidance System

2026-06-15

AI Summary

This work addresses the challenges of limited computational resources, unstable communication, and high annotation costs in edge scenarios such as rural clubs by proposing an instance-aware knowledge distillation framework. The method integrates domain priors from a teacher model with instance-level knowledge from a foundation model to generate high-quality pseudo-labels, enabling semi-supervised training of a lightweight multi-task student model that simultaneously performs obstacle detection and monocular depth estimation. By innovatively mitigating pseudo-label bias, the approach achieves superior instance segmentation performance over the teacher model and significantly reduced depth estimation error, while operating at 6.46 FPS on edge devices with 22.68× fewer FLOPs and 14.33× fewer parameters. Its effectiveness has been validated in real-world deployments.

Abstract

Collision avoidance systems have evolved toward camera-based deep learning approaches for driving scene understanding. However, deployment in edge environments such as country clubs is constrained by limited computational resources and unreliable communication infrastructure. Moreover, constructing large-scale datasets for the target domain involves substantial annotation cost. To address these limitations, we propose an instance-aware knowledge distillation framework for semi-supervised learning. Specifically, we generate pseudo labels that mitigate teacher bias by leveraging domain priors from the teacher and instance-centric knowledge from foundation models. The trained lightweight student is deployed in the proposed collision avoidance system and performs multiple dense prediction tasks in real-time. The system detects frontal obstacles and encodes their spatial information into controller area network messages for automated guided vehicle operation. To achieve this, we construct a large-scale country club dataset and perform field validation of the proposed system. Experimental results demonstrate that the student outperforms the large teacher in instance segmentation while mitigating performance degradation in monocular depth estimation. Compared with the teacher, the student reduces FLOPs by 22.68$\times$ and parameters by 14.33$\times$, achieving 6.46 FPS on a low-cost edge device.

Problem

• collision avoidance system
• edge deployment
• semi-supervised learning
• annotation cost
• computational constraints

Innovation

• instance-aware knowledge distillation
• semi-supervised learning
• multi-task dense prediction
• edge deployment
• pseudo-labeling with foundation models