A newly developed multi-scenario dataset is introduced for the detection and tracking of road vehicles and pedestrians. The dataset, captured through high-resolution surveillance cameras in various traffic environments such as highways, toll stations, gas stations, and urban intersections, consists of 34,383 images for training and 1,598 images for validation. It covers five major object categories-car, bus, truck, bike, and person-captured under diverse lighting and weather conditions to improve its applicability in real-world settings. Furthermore, to address different traffic density scenarios, the dataset is categorized into three subsets: low, medium, and high density. This categorization facilitates comprehensive experiments on four object detection algorithms, ensuring robustness and accuracy across varied conditions. Experimental results demonstrate superior performance, highlighting the dataset's potential as a foundational resource for intelligent transportation systems (ITS) research.
Construction and Performance Evaluation of a Comprehensive Multi-Scenario Road Vehicle and Pedestrian Detection Dataset
15.11.2024
8298747 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Vehicle-road cooperation system and road pedestrian detection method
Europäisches Patentamt | 2021
|Pedestrian-vehicle safety early warning device for pedestrian-vehicle intersection road section
Europäisches Patentamt | 2024
|