Flickr1024

随着双摄像头在最近发布的智能手机中的普及，越来越多的超分辨率 (SR) 方法被提出来提高立体图像对的分辨率。但是，缺乏高质量的立体声数据集限制了该领域的研究。为了便于训练和评估新颖的立体SR算法，本文提出了一个名为Flickr1024的大规模立体数据集，该数据集包含1024对高质量图像，并涵盖了各种场景。我们首先介绍数据采集和处理管道，然后比较几种流行的立体声数据集。最后，我们进行跨数据集实验，以调查数据集带来的潜在好处。实验结果表明，与KITTI和Middlebury数据集相比，我们的Flickr1024数据集可以帮助处理过拟合问题，并显着提高立体声SR方法的性能。With the widespread adoption of dual cameras in recently released smartphones, an increasing number of super-resolution (SR) methods have been proposed to enhance the resolution of stereo image pairs. However, the lack of high-quality stereo datasets has limited research progress in this field. To facilitate the training and evaluation of novel stereo SR algorithms, this paper proposes a large-scale stereo dataset named Flickr1024, which contains 1024 pairs of high-quality images covering diverse scenarios. We first introduce the data acquisition and processing pipeline, followed by a comparison of several popular stereo datasets. Finally, we conduct cross-dataset experiments to investigate the potential benefits brought by the dataset. The experimental results show that, compared with the KITTI and Middlebury datasets, our Flickr1024 dataset can help alleviate overfitting issues and significantly improve the performance of stereo SR methods.

搜集汇总

背景与挑战

背景概述

Flickr1024是一个大规模立体图像数据集，包含1024对高质量图像，涵盖多种场景，旨在支持立体超分辨率算法的训练与评估，以提升性能并减少过拟合。该数据集由国防科技大学于2019年发布，相关论文和资源可通过公开链接获取。