TinyFace

该数据集名为TinyFace，是从公共网络数据中收集的低分辨率人脸数据集，主要用于1:N识别测试。该数据集的特点是图像在不受控制的姿态、光照、遮挡和背景条件下收集。此外，它专注于低分辨率人脸识别的挑战，规模上包含了适用于1:N识别的低分辨率图像。该数据集的主要任务是人脸识别。

The dataset named TinyFace is a low-resolution face dataset collected from public web data, primarily used for 1:N recognition testing. It is characterized by images captured under uncontrolled conditions including pose, illumination, occlusion, and background. Furthermore, it focuses on the challenges of low-resolution face recognition, and includes low-resolution images suitable for 1:N recognition in its corpus. The primary task of this dataset is face recognition.

搜集汇总

背景与挑战

背景概述

TinyFace是一个专注于原生低分辨率人脸识别的大规模基准数据集，包含5,139个身份和169,403张平均20×16像素的低分辨率人脸图像，所有图像均采集自真实网络环境下的多样化场景。该数据集首次系统性地研究了真实低分辨率条件下的人脸识别问题，区别于传统人工降采样的研究方法。