零样本出入境证件智能鉴伪识别数据

本数据集用于支撑出入境证件防伪特征识别算法在零样本条件下的实验验证与性能评估。在实际证件鉴伪应用中，由于各国证件种类繁多且更新频繁，新版本证件样本往往难以及时收集，导致传统依赖大量标注样本训练的识别方法难以适应新类别证件。为解决这一问题，本项目开展了零样本证件鉴伪算法研究，并构建了用于算法验证的多光谱证件图像测试数据。本数据集共包含20组出入境证件图像样本，其中包括10组真实证件样本和10组伪假证件样本。每组样本均包含白光、红外和紫外三种光谱条件下的图像，用于模拟真实证件在不同光谱成像条件下的防伪特征表现。通过多模态图像数据，可以更全面地反映证件在不同光谱下的纹理、荧光及材料特征，从而为证件真伪识别算法提供多源信息支持。数据格式为JPG，数据量43.7MB。

This dataset is designed to support the experimental validation and performance evaluation of anti-counterfeiting feature recognition algorithms for entry-exit travel documents under zero-shot conditions. In real-world document authentication applications, given the vast variety of travel documents across countries and their frequent updates, samples of new-version documents are often not collected in a timely manner. This renders traditional recognition methods that rely on large-scale labeled training samples incapable of adapting to new document categories. To tackle this challenge, this project carried out research on zero-shot document anti-counterfeiting algorithms and built multi-spectral document image test data for algorithm validation. This dataset consists of 20 sets of entry-exit travel document image samples, among which 10 sets are genuine document samples and 10 sets are counterfeit document samples. Each sample set includes images captured under three spectral modes: white light, infrared, and ultraviolet, aiming to simulate the anti-counterfeiting feature manifestations of real documents under different spectral imaging conditions. By leveraging multimodal image data, the texture, fluorescence, and material characteristics of documents across different spectra can be comprehensively captured, thereby providing multi-source information support for document authenticity recognition algorithms. The data is stored in JPG format, with a total size of 43.7 MB.

搜集汇总

背景与挑战

背景概述

该数据集用于支持零样本条件下出入境证件防伪特征识别算法的验证与评估，旨在应对新证件样本收集困难的挑战。它包含20组证件图像样本，涵盖真实和伪假类别，每组均提供白光、红外和紫外三种光谱图像，以全面反映防伪特征，数据格式为JPG，总量约44MB。