"SCPFL-SSE_Data"

"MNIST (Modified National Institute of Standards and Technology) is a classic benchmark dataset for handwritten digit recognition. It consists of 70,000 grayscale images of size 28\u00d728 pixels, split into 60,000 training and 10,000 test samples, covering digits 0 through 9. Despite its widespread use, MNIST is considered relatively simple and no longer challenging for modern models.Fashion-MNIST is a direct drop-in replacement for MNIST, designed to be more challenging. It also contains 70,000 28\u00d728 grayscale images, with the same training\/test split, but instead of digits, it features ten categories of clothing items (e.g., T-shirt, trouser, dress, sneaker). Fashion-MNIST tests a model\u2019s ability to handle more complex patterns while keeping the same data format and size.CIFAR-10 (Canadian Institute for Advanced Research) is a more realistic and difficult dataset of natural images. It contains 60,000 32\u00d732 color RGB images, divided into 50,000 training and 10,000 test images across ten mutually exclusive classes: airplanes, automobiles, birds, cats, deer, dogs, frogs, horses, ships, and trucks. The low resolution and intra-class variability make it a standard benchmark for image classification.SVHN (Street View House Numbers) is a real-world dataset obtained from Google Street View images. It contains 600,000 32\u00d732 color images of house numbers (digits 0\u20139), with a training set of about 73,000 digits, an additional 531,000 \u201cextra\u201d training samples, and a test set of 26,032. Unlike MNIST, SVHN includes distracting digits and varying backgrounds, making digit recognition more challenging and closer to real-world scenarios.These four datasets collectively span handwritten digits, fashion items, natural objects, and street\u2011view digits, providing a gradient of difficulty for evaluating computer vision and machine learning models."