Multi-PIE

Multi-PIE (Multi Pose, Illumination, Expressions) 数据集由 337 位受试者在不同姿势、光照和表情下拍摄的人脸图像组成。姿势范围包含 15 个离散视图，捕获面部轮廓到轮廓。使用位于房间不同位置的 19 个手电筒模拟照明变化。

The Multi-PIE (Multi Pose, Illumination, Expressions) dataset comprises facial images acquired from 337 subjects under diverse poses, illumination conditions and facial expressions. The pose range includes 15 discrete viewpoints, capturing the full span from one facial profile to another. Nineteen flashlights placed at different locations within the experimental room were used to simulate illumination variations.

搜集汇总

背景与挑战

背景概述

Multi-PIE是一个人脸图像数据集，包含337位受试者在15种不同姿势、19种光照变化和多种表情下的图像，主要用于人像重新照明等计算机视觉任务。该数据集由卡内基梅隆大学于2008年发布，支持面部识别和分析研究。