LTV-YOLO: A Lightweight Thermal Object Detector for Young Pedestrians in Adverse Conditions

TL;DR

This work tackles the challenge of detecting young vulnerable road users (VRUs) under adverse lighting and weather by leveraging long-wave infrared (LWIR) thermal imaging. It introduces LTV-YOLO, a lightweight, thermal-only detector based on YOLO principles, featuring depthwise separable convolutions and a feature pyramid network to enable accurate small-object detection on edge devices. The model is trained with a composite loss emphasizing localization accuracy and employs edge-optimized preprocessing, achieving real-time performance on Jetson platforms with strong mAP on public and custom VRU datasets while maintaining a compact footprint (~1.6M parameters). The results demonstrate substantial practical impact for safety‑critical applications in smart cities, school zones, and autonomous systems, while highlighting the need for more child-focused thermal datasets and potential sensor-fusion enhancements for robustness.

Abstract

Detecting vulnerable road users (VRUs), particularly children and adolescents, in low light and adverse weather conditions remains a critical challenge in computer vision, surveillance, and autonomous vehicle systems. This paper presents a purpose-built lightweight object detection model designed to identify young pedestrians in various environmental scenarios. To address these challenges, our approach leverages thermal imaging from long-wave infrared (LWIR) cameras, which enhances detection reliability in conditions where traditional RGB cameras operating in the visible spectrum fail. Based on the YOLO11 architecture and customized for thermal detection, our model, termed LTV-YOLO (Lightweight Thermal Vision YOLO), is optimized for computational efficiency, accuracy and real-time performance on edge devices. By integrating separable convolutions in depth and a feature pyramid network (FPN), LTV-YOLO achieves strong performance in detecting small-scale, partially occluded, and thermally distinct VRUs while maintaining a compact architecture. This work contributes a practical and scalable solution to improve pedestrian safety in intelligent transportation systems, particularly in school zones, autonomous navigation, and smart city infrastructure. Unlike prior thermal detectors, our contribution is task-specific: a thermally only edge-capable design designed for young and small VRUs (children and distant adults). Although FPN and depthwise separable convolutions are standard components, their integration into a thermal-only pipeline optimized for short/occluded VRUs under adverse conditions is, to the best of our knowledge, novel.

Figures (6)

• Figure 1: Thermal detection pipeline of LTV-YOLO. The system receives input from a thermal camera, processes it through lightweight convolution and depthwise blocks, merges multi-scale features via FPN, and outputs bounding boxes or alerts.
• Figure 2: Sensor module for thermal data collection: ZED2i (top) and FLIR ADK (bottom).
• Figure 3: Detection confidence over time: original vs. downsized video.
• Figure 4: Detection consistency over time: original vs. downsized video.
• Figure 5: Confidence score distribution: original vs. downsized video.