MIDV-2020

MIDV-2020数据集由俄罗斯科学院计算机科学与控制联邦研究中心创建，包含1000种不同的模拟身份文档，每种文档都有独特的文本字段值和人工生成的面部图像。数据集包括1000个视频片段、2000张扫描图像和1000张照片，总计72409张标注图像。这些图像涵盖了多种捕捉条件，如光照、背景和文档方向的变化，旨在为身份文档的检测、识别和文本字段识别等任务提供丰富的训练和测试数据。该数据集的应用领域包括自动化数据输入、远程身份验证和反欺诈检测等。

The MIDV-2020 dataset was created by the Federal Research Center for Computer Science and Control of the Russian Academy of Sciences. It contains 1,000 distinct simulated identity documents, each with unique text field values and artificially generated facial images. The dataset includes 1,000 video clips, 2,000 scanned images, and 1,000 photographs, with a total of 72,409 annotated images. These images cover various capture conditions, including variations in lighting, background, and document orientation, and are intended to provide rich training and test data for tasks such as identity document detection, recognition, and text field recognition. The application scenarios of this dataset include automated data entry, remote identity verification, anti-fraud detection, and other related fields.

搜集汇总

背景与挑战

背景概述

MIDV-2020是一个全面的基准数据集，专门用于身份证件分析，由L3i实验室制作和拥有，旨在支持相关研究和开发。