Evaluation of Thermal Imaging on Embedded GPU Platforms for Application in Vehicular Assistance Systems

TL;DR

This paper evaluates thermal object detection performance on embedded GPUs for vehicular systems, introducing a large thermal dataset and optimizing YOLO models for real-time detection in challenging conditions.

Contribution

It presents a new large-scale thermal dataset, trains and optimizes YOLO networks for embedded GPU platforms, and demonstrates significant FPS improvements for automotive applications.

Findings

Achieved 11 fps on Nvidia Jetson Nano

Achieved 60 fps on Nvidia Xavier NX

Validated detection accuracy with multiple metrics

Abstract

This study is focused on evaluating the real-time performance of thermal object detection for smart and safe vehicular systems by deploying the trained networks on GPU & single-board EDGE-GPU computing platforms for onboard automotive sensor suite testing. A novel large-scale thermal dataset comprising of > 35,000 distinct frames is acquired, processed, and open-sourced in challenging weather and environmental scenarios. The dataset is a recorded from lost-cost yet effective uncooled LWIR thermal camera, mounted stand-alone and on an electric vehicle to minimize mechanical vibrations. State-of-the-art YOLO-V5 networks variants are trained using four different public datasets as well newly acquired local dataset for optimal generalization of DNN by employing SGD optimizer. The effectiveness of trained networks is validated on extensive test data using various quantitative metrics which include precision, recall curve, mean average precision, and frames per second. The smaller network variant of YOLO is further optimized using TensorRT inference accelerator to explicitly boost the frames per second rate. Optimized network engine increases the frames per second rate by 3.5 times when testing on low power edge devices thus achieving 11 fps on Nvidia Jetson Nano and 60 fps on Nvidia Xavier NX development boards.