IBims-1 (Independent benchmark images and matched scans v1)

iBims-1（独立基准图像和匹配扫描 - 版本 1）是一个新的高质量 RGB-D 数据集，专为测试单图像深度估计 (SIDE) 方法而设计。定制的采集设置由数字单镜头反光 (DSLR) 相机和高精度激光扫描仪组成，用于获取各种室内场景的高分辨率图像和高精度深度图。与相关的 RGB-D 数据集相比，iBims-1 因其非常低的噪声水平、锐利的深度过渡、无遮挡和高深度范围而脱颖而出。我们的数据集由以下部分组成： 核心数据集：100 个 RGB-D 图像对，包含高分辨率和低分辨率的各种室内场景 用于无效、透明和平面区域（桌子、地板、墙壁）的遮罩 用于不同深度转换的遮罩 相机校准参数辅助数据集： - 核心数据集的每个图像的 56 种不同颜色和几何增强 - 用于测试 MVS 方法的其他手持图像 - 打印图案的图像和张贴在墙上的照片以评估纹理平面表面的性能 - 几个 RGB-D具有不同照明的静态场景的图像序列

iBims-1 (Independent Benchmark Images and Matched Scans - Version 1) is a novel high-quality RGB-D dataset specifically developed for testing single-image depth estimation (SIDE) methods. The custom acquisition setup, composed of a digital single-lens reflex (DSLR) camera and a high-precision laser scanner, is utilized to capture high-resolution images and high-precision depth maps of diverse indoor scenes. Compared with existing related RGB-D datasets, iBims-1 stands out for its extremely low noise level, sharp depth transitions, absence of occlusions, and wide depth range. Our dataset comprises the following components:
1. Core Dataset: 100 RGB-D image pairs covering various indoor scenes with both high-resolution and low-resolution variants; masks for invalid, transparent, and planar regions (tables, floors, walls); masks for different depth transitions; and camera calibration parameters.
2. Auxiliary Dataset:
- 56 distinct color and geometric augmentations for each image in the core dataset;
- Additional handheld images for testing multi-view stereo (MVS) methods;
- Images of printed patterns and wall-mounted photographs to evaluate performance on textured planar surfaces;
- Several RGB-D image sequences of static scenes under different illumination conditions.

搜集汇总

背景与挑战

背景概述

IBims-1是一个高质量RGB-D数据集，专为测试单图像深度估计方法而设计，通过DSLR相机和激光扫描仪采集，具有低噪声、锐利深度过渡和高深度范围等特点。核心数据集包含100个室内场景的RGB-D图像对，并提供遮罩和辅助数据，适用于计算机视觉研究。该数据集由慕尼黑工业大学于2018年发布，相关论文发表于ECCV研讨会。