EarSegDB 25

This dataset is created for the purpose of automatic segmentation of the human ear in an unconstrained environment. It consists of 1275 ear images of both left and right ear from 25 different persons and 1275 manually created segmentation masks for each respective raw ear image. The raw image data is divided into two parts- Train and Test. Four images from each person (two from right ear and two from left ear) are selected randomly and kept in the Test folder (100 images). The rests are kept in Train folder (1175 images). The masks are also divided in the same way as the respective raw ear images and kept in two folders as Trainannot (1175 images) and Testannot (100 images).
The collected data is completely original, new and the samples are collected from students, teachers and employees of the Gauhati University, Guwahati, Asssam, India and one of its affiliated colleges (Pub Kamrup College, Baihata Chariali, Kamrup, Assam India). Data is collected from both Male and Female of age range 21-58 years using a smartphone camera with varying the angle and distance and having no control over the environment. The masks are manually created for each of the images keeping the ear part as white and the non-ear part as black.
A Table with title "EarSegDB 25 info" is also there which carries additional information such as left and right ear images, age and gender of each person. This additional data could be useful for researchers to carry out further works such as age and gender prediction etc. using ear images.

本数据集旨在实现无约束环境下的人耳自动分割任务。数据集包含25名不同受试者的左右耳共1275张原始耳部图像，以及与每张原始图像一一对应的1275张人工标注分割掩码。原始图像数据划分为训练集（Train）与测试集（Test）两个子集：每名受试者随机选取4张耳部图像（右耳2张、左耳2张），总计100张图像划入测试集文件夹；剩余1175张图像归入训练集文件夹。分割掩码的划分规则与原始图像完全一致，分别存储于Trainannot（1175张）与Testannot（100张）两个标注文件夹中。本次采集的数据均为原创全新样本，采集对象为印度阿萨姆邦古瓦哈提大学（Gauhati University, Guwahati, Assam, India）及其附属学院普布卡姆鲁普学院（Pub Kamrup College, Baihata Chariali, Kamrup, Assam, India）的学生与教职工，涵盖年龄介于21至58岁之间的男性与女性群体。数据采集全程采用智能手机摄像头，未对拍摄角度、拍摄距离及拍摄环境进行任何限制。所有图像的分割掩码均为人工标注：耳部区域标注为白色，非耳部区域标注为黑色。本数据集附带一张标题为"EarSegDB 25 info"的表格，其中包含每名受试者的左右耳图像对应信息、年龄及性别等附加元数据。此类附加信息可辅助研究人员开展基于耳部图像的年龄与性别预测等后续研究工作。