Reubencf/street-scene-descriptions

该数据集包含超过10,000个街道级别的图像样本，配有详细的文本场景描述和丰富的元数据，包括地理位置、时间、天气和基础设施细节。每个条目都有一个自然语言标题，描述从车辆视角捕捉的不同光照和道路条件下的城市和住宅环境。数据结构的目的是支持视觉语言任务，提供指南针角度、序列ID和特定道路类型等属性以及生成的描述。

This dataset contains over 10,000 street-level image samples accompanied by detailed textual scene descriptions and rich metadata including geolocation, time, weather, and infrastructure details. Each entry features a natural language caption describing urban and residential environments captured from vehicle perspectives under varying lighting and road conditions. The data is structured to support vision-language tasks, offering attributes like compass angle, sequence IDs, and specific road types alongside the generated descriptions.