High-Resolution Inpainting Dataset for Airborne Object Detection
收藏资源简介:
该数据集由Spleenlab GmbH与伊尔梅瑙理工大学的研究团队创建,旨在为无人机自主飞行中的安全关键对象检测提供支持。数据集包含7000张高分辨率图像,通过生成模型(如Pix2Pix和Stable Diffusion)进行图像修复生成,涵盖了不同天气、云层和光照条件下的背景。数据集的内容包括多种空中物体,如小型飞机、直升机、无人机等,每个图像都带有标注的边界框和分割掩码。数据集的创建过程基于现有的背景图像,通过修复技术插入目标物体,生成逼真的合成图像。该数据集主要用于训练和验证无人机检测与避让系统中的对象检测模型,旨在提高无人机在复杂环境中的安全性和可靠性。
This dataset was developed by Spleenlab GmbH and a research team from Technische Universität Ilmenau, aiming to support safety-critical object detection for autonomous unmanned aerial vehicle (UAV) flight. It contains 7000 high-resolution images generated via image inpainting using generative models such as Pix2Pix and Stable Diffusion, with backgrounds covering various weather, cloud cover and lighting conditions. The dataset includes multiple aerial targets including small aircraft, helicopters, UAVs and others, and each image is equipped with annotated bounding boxes and segmentation masks. The dataset was constructed based on existing background images, where target objects are inserted through inpainting techniques to produce realistic synthetic images. This dataset is primarily used to train and validate object detection models in UAV detection and avoidance systems, with the goal of improving the safety and reliability of UAVs in complex environments.




