projectsidewalk/rampnet-crop-model-dataset

RampNet是一个两阶段的数据集，通过使用来自三个美国城市（纽约市、波特兰、本德）的政府位置数据自动生成超过210,000个注释的Google Street View全景图，用于训练路缘坡检测模型。该数据集包括手动标记的小型裁剪集、项目人行道裁剪预训练集、第一阶段生成的数据集和一个手动标记的地面真实集。

RampNet is a two-stage dataset that uses government location data from three US cities (New York City, Portland, Bend) to automatically generate over 210,000 annotated Google Street View panoramas for training curb ramp detection models. The dataset includes a manually labeled crop set, a Project Sidewalk crop pre-training set, a dataset generated in the first stage, and a manually labeled ground truth set.