RealToF
收藏资源简介:
RealToF数据集是由中南大学和清华大学等机构创建的,旨在支持移动设备上的ToF和双目立体视觉深度感知研究。该数据集包含150个多样化的室内场景,使用华为P40Pro设备上的ToF和双RGB摄像头采集数据,并通过Intel RealSense深度摄像头和激光测距仪获取精确的地面真实数据。数据集的创建过程结合了物理原理和环境因素,确保了深度误差的准确表征。RealToF数据集主要应用于3D重建和3D分割等下游任务,旨在提高移动设备上的深度感知精度。
The RealToF Dataset was developed by institutions including Central South University and Tsinghua University, aiming to support research on ToF and binocular stereo vision-based depth perception on mobile devices. It comprises 150 diverse indoor scenes, with data collected using the ToF and dual RGB cameras on a Huawei P40 Pro device, while precise ground-truth data was acquired via an Intel RealSense depth camera and a laser rangefinder. The development of the RealToF Dataset integrates physical principles and environmental factors to ensure accurate characterization of depth errors. Primarily utilized for downstream tasks such as 3D reconstruction and 3D segmentation, the RealToF Dataset aims to improve the accuracy of depth perception on mobile devices.

- 1MobiFuse: A High-Precision On-device Depth Perception System with Multi-Data Fusion中南大学、清华大学、上海传音控股有限公司、首尔国立大学 · 2024年



