USPS 手写数字数据集

USPS 全称 United States Postal Service，这是一个数字识别的数据集，数据由美国邮政署自动扫描信件获得。该数据集共包含了 9,298 个 16×16 像素的灰度样本，其中训练集包含 7,291 个，数据集包含 2,007 个, 图像均格式居中，标准化并显示多种字体。

USPS, short for United States Postal Service, is a handwritten digit recognition dataset. The dataset is collected through automatic scanning of letters by the United States Postal Service. It contains a total of 9,298 grayscale samples with a resolution of 16×16 pixels, where 7,291 samples are allocated to the training set and the remaining 2,007 samples form the rest of the dataset. All images are centered, normalized, and feature diverse fonts.

搜集汇总

背景与挑战

背景概述

USPS手写数字数据集是一个用于数字识别的经典数据集，包含9,298个16×16像素的灰度手写数字图像，来源于美国邮政署的信件扫描。数据集已预先分为7,291个训练样本和2,007个测试样本，图像经过居中、标准化处理，并涵盖多种字体，适用于OCR和图像识别任务。