AIXI-Shape 和 AIXI-Color

本研究介绍了两个新的数据集：AIXI-Shape和AIXI-Color，由巴利阿里群岛大学的UGiVIA研究组创建。这两个数据集包含50000张训练图像和2000张验证图像，每张图像大小为128x128像素。AIXI-Shape包含圆形、方形和十字形等简单图形，而AIXI-Color则包含不同大小和强度的圆形。数据集的创建过程涉及使用特定的生成函数来定义图像中每个像素的重要性。这些数据集主要用于评估解释性人工智能（XAI）技术，特别是在图像处理任务中的应用，旨在解决模型解释性和透明度的问题。

This study introduces two novel datasets: AIXI-Shape and AIXI-Color, created by the UGiVIA Research Group at the University of the Balearic Islands. Both datasets consist of 50,000 training images and 2,000 validation images, with each image having a resolution of 128×128 pixels. AIXI-Shape features simple shapes such as circles, squares, and crosses, while AIXI-Color contains circles of varying sizes and intensities. The creation of these datasets involves using specific generation functions to define the importance of each pixel in the images. These datasets are primarily used to evaluate eXplainable Artificial Intelligence (XAI) techniques, particularly for applications in image processing tasks, with the goal of addressing issues related to model interpretability and transparency.