MS-YOLO: Infrared Object Detection for Edge Deployment via MobileNetV4 and SlideLoss

2025-09-25

AI Summary

This work addresses three key challenges in infrared target detection for urban autonomous driving: severe class imbalance, thermal noise interference, and stringent computational constraints on edge devices. To tackle these, we propose a lightweight and efficient detection framework. Methodologically, we replace YOLOv8’s CSPDarknet backbone with MobileNetV4, reducing computational cost by 1.5% while preserving detection accuracy. Additionally, we introduce SlideLoss—a dynamic loss function that adaptively focuses on hard and occluded samples—to jointly optimize precision and recall. Evaluated on the FLIR ADAS V2 dataset, our model achieves competitive mAP—matching state-of-the-art methods—while consuming only 6.7 GFLOPs. The framework significantly improves the trade-off between detection efficiency and robustness under low-light and adverse weather conditions, demonstrating strong suitability for real-time deployment on resource-constrained edge platforms.

Technology Category

• Computer Vision: Object Detection & Categorization
• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Infrared imaging has emerged as a robust solution for urban object detection under low-light and adverse weather conditions, offering significant advantages over traditional visible-light cameras. However, challenges such as class imbalance, thermal noise, and computational constraints can significantly hinder model performance in practical settings. To address these issues, we evaluate multiple YOLO variants on the FLIR ADAS V2 dataset, ultimately selecting YOLOv8 as our baseline due to its balanced accuracy and efficiency. Building on this foundation, we present exttt{MS-YOLO} ( extbf{M}obileNetv4 and extbf{S}lideLoss based on YOLO), which replaces YOLOv8's CSPDarknet backbone with the more efficient MobileNetV4, reducing computational overhead by extbf{1.5%} while sustaining high accuracy. In addition, we introduce emph{SlideLoss}, a novel loss function that dynamically emphasizes under-represented and occluded samples, boosting precision without sacrificing recall. Experiments on the FLIR ADAS V2 benchmark show that exttt{MS-YOLO} attains competitive mAP and superior precision while operating at only extbf{6.7 GFLOPs}. These results demonstrate that exttt{MS-YOLO} effectively addresses the dual challenge of maintaining high detection quality while minimizing computational costs, making it well-suited for real-time edge deployment in urban environments.

Problem

• Enhancing infrared object detection in low-light urban environments
• Addressing class imbalance and thermal noise in thermal imaging
• Reducing computational costs for real-time edge deployment

Innovation

• Replaces CSPDarknet with efficient MobileNetV4 backbone
• Introduces SlideLoss to emphasize underrepresented samples
• Achieves high precision with low computational cost