Spatiotemporal Object Detection for Improved Aerial Vehicle Detection in Traffic Monitoring

TL;DR

This paper develops a spatiotemporal object detection approach for UAV-based traffic monitoring, introducing a new dataset and enhancements to YOLO with attention mechanisms, leading to significant accuracy improvements.

Contribution

It presents a novel spatiotemporal vehicle detection dataset and improves YOLO-based detection with temporal and attention mechanisms for better UAV traffic monitoring.

Findings

16.22% performance improvement over single frame models

Introduction of the STVD dataset with 6,600 annotated images

Attention mechanisms further enhance detection accuracy

Abstract

This work presents advancements in multi-class vehicle detection using UAV cameras through the development of spatiotemporal object detection models. The study introduces a Spatio-Temporal Vehicle Detection Dataset (STVD) containing 6, 600 annotated sequential frame images captured by UAVs, enabling comprehensive training and evaluation of algorithms for holistic spatiotemporal perception. A YOLO-based object detection algorithm is enhanced to incorporate temporal dynamics, resulting in improved performance over single frame models. The integration of attention mechanisms into spatiotemporal models is shown to further enhance performance. Experimental validation demonstrates significant progress, with the best spatiotemporal model exhibiting a 16.22% improvement over single frame models, while it is demonstrated that attention mechanisms hold the potential for additional performance gains.