Expert, Pitbull, MNIST, FMNIST, CIFAR10, Tiny ImageNet, Stanford Dogs, Random Gaussian, Augmented CIFAR10

本研究使用了九个精心挑选的数据集，每个数据集都针对不同复杂度的分类任务。这些数据集包括图像数据，如MNIST、FMNIST、CIFAR10等，以及两个特殊数据集：一个通过随机选择像素值和类分配来最大化熵，另一个通过对CIFAR10数据集中的图像进行增强来展示高冗余。所有数据集都经过标准化处理，统一分辨率为128x128。这些数据集的应用领域广泛，旨在通过评估数据集质量来提升机器学习模型的性能和准确性。

In this study, nine carefully curated datasets are employed, each tailored for classification tasks with varying complexities. These datasets include image data such as MNIST, FMNIST, CIFAR-10, among others, plus two specialized datasets: one maximizes entropy by randomly selecting pixel values and class labels, while the other demonstrates high redundancy through image augmentation on the CIFAR-10 dataset. All datasets have undergone standardization and been unified to a consistent resolution of 128×128. These datasets cover a wide range of application scenarios, aiming to improve the performance and accuracy of machine learning models by evaluating dataset quality.