Brooklyn

Brooklyn is a dataset for cross-view semantic segmentation, comprising 7,600 images in the training set and 1,900 images in the validation set. The Brooklyn dataset includes satellite images and street-view images with depth map , both of which are well center-aligned. It supports fine-grained building attribute semantic segmentation (land use) and floor recognition (floor) tasks. This paper has been accepted by CVPR 2024.

布鲁克林（Brooklyn）数据集是一款面向跨视图语义分割任务的数据集，其训练集包含7600张图像，验证集包含1900张图像。该数据集涵盖配准精度优异的卫星图像与带有深度图的街景图像，二者均经过良好的中心对齐。本数据集支持细粒度建筑属性语义分割（土地利用）及楼层识别（floor）任务。本论文已被CVPR 2024接收。

搜集汇总

背景与挑战

背景概述

Brooklyn是一个跨视角语义分割数据集，包含卫星和街景图像（带深度图），支持土地利用和楼层识别任务，已被CVPR 2024接受。