AVOIDDS
收藏资源简介:
AVOIDDS是由斯坦福大学开发的一个用于视觉基础的飞机入侵检测的数据集和模拟器,旨在解决飞机碰撞避免问题。该数据集包含72,000张具有各种光照、天气、相对几何和地理位置的逼真图像,用于训练和测试视觉基础的飞机检测模型。数据集创建过程中使用了X-Plane 11飞行模拟器,自动生成边界框标签,无需手动标注。AVOIDDS不仅用于模型训练,还提供了一个闭环模拟器来评估模型在碰撞避免任务中的性能,适用于高风险应用中机器学习系统的鲁棒性设计。
AVOIDDS is a dataset and simulator for vision-based aircraft intrusion detection developed by Stanford University, designed to tackle the aircraft collision avoidance problem. This dataset comprises 72,000 photorealistic images covering diverse lighting conditions, weather scenarios, relative geometries and geographic locations, which are used for training and testing vision-based aircraft detection models. The X-Plane 11 flight simulator was employed during the dataset creation process to automatically generate bounding box labels, eliminating the need for manual annotation. Beyond model training, AVOIDDS also provides a closed-loop simulator to evaluate model performance in collision avoidance tasks, making it suitable for the robustness design of machine learning systems in high-risk applications.
- 1AVOIDDS: Aircraft Vision-based Intruder Detection Dataset and Simulator斯坦福大学 · 2023年



