HouseExpo

该数据集是从SUNCG数据集中生成的包含大规模2D室内布局图像。数据集中还包括由伪SLAM模拟器生成的地图图像，所有图像均被调整为64x64像素的大小。具体规模包括5000张训练图像、200张验证图像以及100张测试图像。该数据集的任务是用于地图补全以及对SLAM方法的评估。

This dataset is generated from the SUNCG dataset and contains large-scale 2D indoor layout images. It also includes map images generated by a pseudo-SLAM simulator, and all images have been resized to 64x64 pixels. Specifically, the dataset consists of 5000 training images, 200 validation images, and 100 test images. The tasks supported by this dataset include map completion and evaluation of SLAM methods.

搜集汇总

背景与挑战

背景概述

HouseExpo是一个基于SUNCG构建的大规模2D室内布局数据集，包含35,126个平面图和252,550个带类别标签的房间，专为移动机器人学习算法设计。该数据集与PseudoSLAM模拟器集成，支持探索和导航任务，数据以JSON格式存储，包括房间统计和可视化功能。