Semantic labeling for improved vehicle detection in aerial imagery

摘要: Growing cities and increasing traffic densities result in an increased demand for applications such as monitoring, analysis, support of rescue work. These share the need accurate detection relevant vehicles, e.g. aerial imagery. Recently, application deep learning based frameworks like Faster R-CNN clearly outperformed conventional methods vehicle images. In this paper, we propose a framework that fuses semantic labeling to integrate contextual information. We achieve improved performance by decreasing number false positive detections while candidate regions classify is reduced. To demonstrate generalization our approach, evaluate various ground sampling distances on publicly available dataset.