SynDroneVision
收藏资源简介:
SynDroneVision是由德国航空航天中心(DLR)与达姆施塔特应用科学大学联合创建的合成数据集,旨在为监控应用中的RGB基础无人机检测提供训练基础。该数据集通过游戏引擎模拟生成,包含多样化的背景、光照条件和无人机模型,为深度学习算法的训练提供了全面的数据支持。SynDroneVision在规模和多样性上均表现出色,包含131,307张标注图像,覆盖多种环境和无人机类型,有效提升了模型性能和鲁棒性,同时显著降低了实际数据采集的时间和成本。该数据集的发布,为无人机检测领域提供了宝贵的资源,尤其在结合实际数据训练时,能够显著提高检测模型的准确性。
SynDroneVision is a synthetic dataset jointly developed by the German Aerospace Center (DLR) and Darmstadt University of Applied Sciences, intended to serve as a training foundation for RGB-based drone detection in surveillance applications. Generated through game engine simulations, the dataset features diverse backgrounds, lighting conditions and drone models, providing comprehensive data support for the training of deep learning algorithms. Excelling in both scale and diversity, SynDroneVision contains 131,307 annotated images covering multiple environments and drone types, which effectively improves model performance and robustness while significantly reducing the time and cost of real-world data collection. The release of this dataset offers a valuable resource for the drone detection field, and when combined with real-world data during training, it can notably enhance the accuracy of detection models.




