Navigation Benchmark Video Dataset for Wearable Edge-AI Thermal Evaluation

This repository contains the experimental artifacts used in the study “Airflow-Optimized Enclosures for Thermal Stability in Wearable Edge-AI Systems.” The dataset includes the navigation video input used for inference benchmarking and the accompanying code used to reproduce the thermal and performance experiments.The benchmark video consists of a 2,850-frame prerecorded navigation sequence (resolution: 1467 × 837 pixels, 30 FPS) used as a consistent workload for evaluating real-time object detection on an NVIDIA Jetson AGX Orin platform running a YOLOv5m model exported to ONNX and deployed with TensorRT.The repository also provides scripts for:running inference using the YOLOv5m ONNX modellogging GPU and CPU telemetry via tegrastatscomputing inference latency and throughput metricsreproducing the temperature and performance analyses reported in the paperThese resources support reproducibility of the airflow-optimized wearable enclosure experiments comparing passive and active cooling configurations.The artifacts are released under a CC-BY license to support open research and reproducibility in edge AI, embedded systems, and assistive navigation technologies.