LingBot-Depth-Dataset

LingBot-Depth-Dataset 是目前社区中规模最大的真实场景 RGB-D 数据集，包含 300 万对高质量样本，其中 200 万对来自真实场景采集、100 万对由渲染生成。每条样本均包含 RGB 图像、传感器原始深度图和真值深度图，可直接用于深度估计与深度补全任务的训练和评估，总规模达 2.71TB。数据集覆盖 Orbbec 335、335L 以及 Intel RealSense D405、D415、D435、D455 共 6 款主流深度相机，将有助于提升模型在不同设备和场景下的适应能力。LingBot-Depth-Dataset is the largest real-world RGB-D dataset currently available in the community. It comprises 3 million high-quality sample pairs, among which 2 million pairs are collected from real-world scenarios and 1 million pairs are synthetically rendered. Each sample pair contains an RGB image, the sensor's raw depth map, and the ground-truth depth map, and can be directly utilized for training and evaluating depth estimation and depth completion tasks. The total size of the dataset amounts to 2.71 TB. Covering a total of 6 mainstream depth camera models including Orbbec 335, 335L, Intel RealSense D405, D415, D435, and D455, this dataset will help improve the adaptability of models across different devices and scenarios.

搜集汇总

背景与挑战

背景概述

LingBot-Depth-Dataset是一个自整理的RGB-D数据集，用于训练深度估计模型LingBot-Depth，每个样本包含RGB图像、原始传感器深度和真实深度。该数据集由真实世界室内场景、机器人操作任务和模拟数据组成，总计约300万样本。