多视角数据集
收藏资源简介:
本文介绍了一个新的多视角数据集,该数据集使用自定义的密集线性相机阵列构建,旨在促进帧插值和视图合成技术之间的公平比较。该数据集包括真实和合成的场景,可以用于评估和训练算法,以生成新视角的图像。数据集由9个Raspberry Pi单元组成,每个单元配备一个12.3万像素的索尼IMX477传感器,以实现高分辨率图像捕获。数据集包含2140帧图像,其中包括10个对象的图像,这些对象在不同的距离上被捕获,以创建不同的视差效果。该数据集可用于评估和训练帧插值和视图合成算法,以生成新视角的图像。
This paper introduces a novel multi-view dataset constructed with a custom dense linear camera array, which aims to facilitate fair comparisons between frame interpolation and view synthesis techniques. This dataset includes both real and synthetic scenes, and can be used to evaluate and train algorithms for generating novel-view images. The dataset consists of 9 Raspberry Pi units, each equipped with a 123,000-pixel Sony IMX477 sensor to enable high-resolution image capture. The dataset contains 2140 image frames, including images of 10 objects captured at different distances to create various parallax effects. This dataset can be used to evaluate and train frame interpolation and view synthesis algorithms for generating novel-view images.

- 1A new dataset and comparison for multi-camera frame synthesis都柏林圣三一学院电子与电气工程系 · 2025年



