背景

Human Parse指将在图像中的人分割为多个语义上一致的区域。
目前该领域主要有以下开源数据集：PASCAL-Person-Part、ATR、LIP、MHP、CIHP。

PASCAL-Person-Part

PASCAL-Person-Part在PSACAL VOC2010数据集基础上增加分割标注，该数据集存在多人场景，其中训练集1716张，测试集1817张，共标注7个类别：‘Background’, ‘Head’, ‘Torso’, ‘Upper Arms’, ‘Lower Arms’, ‘Upper Legs’, ‘Lower Legs’。如下图所示。

下载链接
论文：《Detect What You Can: Detecting and Representing Objects using Holistic Models and Body Parts》

ATR

ATR(Active Template Regression)数据集为单人场景人体解析数据集，该数据集共17700张，其中训练集16000张，验证集700张，测试集1000张。共标注18个类别：‘Background’, ‘Hat’, ‘Hair’, ‘Sunglasses’, ‘Upper-clothes’, ‘Skirt’, ‘Pants’, ‘Dress’, ‘Belt’, ‘Left-shoe’, ‘Right-shoe’, ‘Face’, ‘Left-leg’, ‘Right-leg’, ‘Left-arm’, ‘Right-arm’, ‘Bag’, ‘Scarf’.如下图所示。

下载链接
论文：《Deep Human Parsing with Active Template Regression》

LIP

LIP(Look into Person)数据集为单人场景人体解析数据集，共50462张，其中训练集30462张，验证集10000张，测试集10000张，共标注20个类别：‘Background’, ‘Hat’, ‘Hair’, ‘Glove’, ‘Sunglasses’, ‘Upper-clothes’, ‘Dress’, ‘Coat’, ‘Socks’, ‘Pants’, ‘Jumpsuits’, ‘Scarf’, ‘Skirt’, ‘Face’, ‘Left-arm’, ‘Right-arm’, ‘Left-leg’, ‘Right-leg’, ‘Left-shoe’, ‘Right-shoe’.如下图所示。

下载链接
论文： 《Look into Person: Self-supervised Structure-sensitive Learning and A New Benchmark for Human Parsing》

MHP

MHP（Multi-Human-Parsing）数据集为多人场景人体解析数据集。MHP v1.0包括4980张图片，平均每张图包括3人，训练集3000张，验证集1000张，测试集980张，共标注19个类别：“background”, “hat”, “hair”, “sunglasses”, “upper clothes”, “skirt”, “pants”, “dress”, “belt”, “left shoe”, “right shoe”, “face”, “left leg”, “right leg”, “left arm”, “right arm”, “bag”, “scarf” and “torso skin”，如下图所示：

下载地址：baidu drive (password: cmtp).

MHP v2.0包括25403图片，训练集15403张，验证集5000张，测试集5000张，共标注59个类别： “background”, “cap/hat”, “helmet”, “face”, “hair”, “left-arm”, “right-arm”, “left-hand”, “right-hand”, “protector”, “bikini/bra”, “jacket/windbreaker/hoodie”, “t-shirt”, “polo-shirt”, “sweater”, “singlet”, “torso-skin”, “pants”, “shorts/swim-shorts”, “skirt”, “stockings”, “socks”, “left-boot”, “right-boot”, “left-shoe”, “right-shoe”, “left-highheel”, “right-highheel”, “left-sandal”, “right-sandal”, “left-leg”, “right-leg”, “left-foot”, “right-foot”, “coat”, “dress”, “robe”, “jumpsuit”, “other-full-body-clothes”, “headwear”, “backpack”, “ball”, “bats”, “belt”, “bottle”, “carrybag”, “cases”, “sunglasses”, “eyewear”, “glove”, “scarf”, “umbrella”, “wallet/purse”, “watch”, “wristband”, “tie”, “other-accessary”, “other-upper-body-clothes” and “other-lower-body-clothes”.如下图所示，

下载地址：google drive
github：https://github.com/ZhaoJ9014/Multi-Human-Parsing

CIHP

CIHP（Crowd Instance-level Human Parsing）数据集为多人场景人体解析数据集，共38280张，其中训练集28280张，验证集5000张，测试集5000张，共标注20个类别：Background, Hat, Hair, Sunglasses, Upper-clothes, Dress, Coat, Socks, Pants, Gloves, Scarf, Skirt, Torsoskin, Face, Right/Left arm, Right/Left leg, and Right/Left shoe.如下图所示，

下载地址
论文：《Instance-level Human Parsing via Part Grouping Network》

总结