Computational Imaging 计算成像

数据基于我用Lytro Illum相机拍摄的图像，这些图像已导出为图像数据和深度图。想法是制造和构建用于查看Lytro Image数据并改善结果的工具。可用于：建立一个可从2D RGB图像自动生成深度信息的神经网络；构建工具以查找深度图像中的间隙或孔并自动修复它们；建立一个可以从RGBD图像重建3D像素数据的神经网络。

This dataset is based on images captured using a Lytro Illum camera, which have been exported as image data and depth maps. The core objective is to develop and build tools for viewing Lytro image data and enhancing the resulting outputs. This dataset can be applied to the following tasks: 1. Developing neural networks that automatically generate depth information from 2D RGB images; 2. Building tools to detect gaps or holes in depth maps and perform automatic repairs on them; 3. Developing neural networks that can reconstruct 3D pixel data from RGBD images.

搜集汇总

背景与挑战

背景概述

该数据集包含由Lytro Illum相机拍摄的图像，已导出为图像数据和深度图，可用于训练神经网络自动生成深度信息、修复深度图像缺陷以及从RGBD图像重建3D数据。此外，还提供来自Lenovo Phab2（Project Tango）的立体声数据集，以点云格式支持3D结构重建。