Floor Plan Visual Navigation (FloNa) Dataset

该数据集是一个大规模的数据集，包含了来自iGibson模拟器的20,214个导航场景，跨越了117个静态室内环境。该数据集旨在评估融合了平面图知识的具身视觉导航。此外，数据集还包含了手动标注的平面图和可导航地图，每个场景通过采样起始点和目标点来收集导航情节。该数据集旨在弥合视觉导航与使用平面图信息进行空间规划之间的差距。其规模覆盖了20,214个导航场景，分布在117个不同的环境中，任务重点在于利用平面图进行具身视觉导航。

This is a large-scale dataset containing 20,214 navigation scenarios sourced from the iGibson simulator, spanning 117 static indoor environments. This dataset is designed to evaluate embodied visual navigation that integrates floor plan knowledge. Additionally, it includes manually annotated floor plans and navigable maps, with navigation episodes collected for each scenario by sampling start and target points. The dataset aims to bridge the gap between visual navigation and spatial planning that relies on floor plan information. Covering 20,214 navigation scenarios across 117 distinct environments, the task of this dataset focuses on embodied visual navigation leveraging floor plan knowledge.