MERL-RAV (MERL Reannotation of AFLW with Visibility)

MERL-RAV（MERL Reannotation of AFLW with Visibility）数据集包含超过 19,000 张人脸图像，包含各种头部姿势。每张脸都被手动标记为 68 个地标的真实位置，以及每个地标是否未被遮挡、自遮挡（由于极端的头部姿势）或外部遮挡的附加信息。这些图像由专业标签人员进行注释，并由三菱电机研究实验室 (MERL) 的研究人员监督。

The MERL-RAV (MERL Reannotation of AFLW with Visibility) dataset contains over 19,000 facial images with diverse head poses. Each face is manually annotated with the ground-truth locations of 68 facial landmarks, alongside additional information specifying whether each landmark is unoccluded, self-occluded (resulting from extreme head poses), or externally occluded. These images were annotated by professional labelers and supervised by researchers at Mitsubishi Electric Research Laboratories (MERL).

搜集汇总

背景与挑战

背景概述

MERL-RAV数据集是对AFLW数据集的重标注版本，包含超过1.9万张涵盖多样头部姿势的人脸图像。每张图像均手动标注了68个地标的真实位置，并提供了地标是否被遮挡的可见性信息，注释工作由专业人员在MERL监督下完成。