BRI3L: BRightness Illusion Image dataset for Identification and Localization of illusory perception

BRI3L数据集是由约翰霍普金斯大学等机构的研究人员创建的，包含22,366张图像，涵盖五种亮度错觉类型：赫尔曼网格、同时亮度对比、白色错觉、网格错觉和诱导光栅错觉。每张图像都附带了二值分割掩码，指示图像中的错觉区域。数据集通过标准的心理物理实验进行验证，旨在通过数据驱动的神经网络方法进行错觉的识别和定位。该数据集的应用领域主要集中在视觉错觉的研究，旨在解决错觉图像的自动识别和错觉区域的精确分割问题。

The BRI3L Dataset was created by researchers from Johns Hopkins University and other institutions. It contains 22,366 images covering five types of brightness illusions: Hermann Grid, Simultaneous Brightness Contrast, White's Illusion, Grid Illusion, and Induced Grating Illusion. Each image is accompanied by a binary segmentation mask that indicates the illusion regions within the image. The dataset was validated via standard psychophysical experiments, and aims to enable data-driven neural network approaches for the recognition and localization of visual illusions. Its application fields mainly focus on visual illusion research, aiming to solve the problems of automatic recognition of illusion images and accurate segmentation of illusion regions.