StreetLearn

StreetLearn数据集由英国伦敦的深度思维公司创建，是一个基于Google街景的交互式视觉环境，用于支持端到端导航研究。该数据集覆盖纽约市和匹兹堡两大区域，包含约56000个高分辨率街景图像，模拟真实世界的街道连接图。数据集创建过程中，使用了Google街景的图像和连接信息，确保了城市规模的真实性和多样性。StreetLearn数据集主要用于解决复杂环境下的目标驱动导航问题，通过模拟真实世界的导航任务，如快递员在城市中的导航，来训练和验证导航算法。

The StreetLearn dataset was created by DeepMind, based in London, UK. It is an interactive visual environment built on Google Street View, designed to support end-to-end navigation research. Covering two major urban areas, New York City and Pittsburgh, the dataset contains approximately 56,000 high-resolution Street View images and simulates real-world street connection graphs. During its development, the dataset leveraged Google Street View images and connection information to guarantee the authenticity and diversity of the urban-scale environments. The StreetLearn dataset is mainly used to address goal-driven navigation problems in complex environments, where it trains and validates navigation algorithms by simulating real-world navigation tasks such as urban courier navigation.