4D Light-Field Dataset for Material Recognition

本研究介绍了首个中等规模的光场图像数据集，专为材料识别设计。该数据集由加州大学伯克利分校和加州大学圣地亚哥分校共同创建，包含12种材料类别，每类100张图像，总计约1200张图像。数据集通过Lytro Illum相机捕捉，每张图像具有376×541的空间分辨率和14×14的视角分辨率，旨在通过多视角信息提升材料识别的准确性。此数据集不仅限于材料识别，还可应用于物体检测、图像分割和视角插值等光场相关研究。

This study presents the first medium-scale light field image dataset specifically designed for material recognition. This dataset was co-developed by the University of California, Berkeley and the University of California, San Diego, consisting of 12 material categories with 100 images per category, totaling approximately 1,200 images. Captured using a Lytro Illum camera, each image has a spatial resolution of 376×541 and an angular resolution of 14×14, aiming to improve the accuracy of material recognition via multi-view information. This dataset is not limited to material recognition, and can also be applied to light field-related research such as object detection, image segmentation, and view interpolation.