Mono2Stereo
收藏资源简介:
Mono2Stereo数据集是一个大规模的高质量立体转换数据集,包含超过2.4百万的立体图像对,涵盖室内外环境、动画和真实生活内容以及不同复杂度的场景。该数据集由200个视频和电影(超过2000万帧)中筛选和裁剪得到,采用侧-by-side 3D格式,每帧图像都有对应的左右视角图像。数据集被分为训练集和测试集,用于评估立体转换模型在视差准确性、场景复杂度和颜色一致性三个关键方面的性能。
The Mono2Stereo Dataset is a large-scale, high-quality stereo conversion dataset containing over 2.4 million stereo image pairs. It covers indoor and outdoor environments, animated and real-life content, as well as scenes with varying complexity. This dataset is curated and cropped from 200 videos and films (over 20 million frames), adopting the side-by-side 3D format, where each stereo image pair includes its corresponding left and right viewpoint images. The dataset is split into a training set and a test set, which is used to evaluate the performance of stereo conversion models across three critical aspects: disparity accuracy, scene complexity, and color consistency.




