∼20 k real-world image dataset
收藏资源简介:
该数据集由Dalle Molle Institute for Artificial Intelligence (IDSIA)创建,包含约20,000张图像,用于训练深度学习模型以实现微型无人机的门检测。图像采集于实验室环境,涵盖不同背景和光照条件,经过模拟器和现实世界的图像增强,以缩小模拟与现实之间的差距。数据集适用于无需地图的门到门视觉导航任务,旨在推动微型无人机在受限环境下的自主导航能力。
This dataset was developed by the Dalle Molle Institute for Artificial Intelligence (IDSIA), comprising approximately 20,000 images for training deep learning models to perform door detection on micro aerial vehicles. The images are captured in laboratory settings, covering diverse backgrounds and lighting conditions, and have undergone image augmentation using both simulated and real-world data to narrow the sim-to-real gap. This dataset is suitable for map-free door-to-door visual navigation tasks, aiming to advance the autonomous navigation capabilities of micro aerial vehicles in constrained environments.




