AHP 去遮挡数据集

AHP 全称 Amodal Human Perception，由 56,599 张图像组成，这些图像来自几个大规模的实例分割和检测数据集 (COCO, VOC (w/SBD), LIP, Objects365, OpenImages) 。每张图像都有一个像素级的完整人类分割掩码。

AHP, whose full name is Amodal Human Perception, consists of 56,599 images sourced from several large-scale instance segmentation and detection datasets including COCO, VOC (w/SBD), LIP, Objects365, and OpenImages. Each image is equipped with a pixel-level complete human segmentation mask.

搜集汇总

背景与挑战

背景概述

AHP去遮挡数据集全称为Amodal Human Perception，包含56,599张来自多个大规模实例分割和检测数据集的图像，每张图像均提供像素级完整人类分割掩码。该数据集专为人类去遮挡任务设计，分为训练集、验证集和测试集，其中训练集包含56,302张图像。