MoLa IR CovSurv

This repository presents one of the datasets described in the article "AI based monitoring of different risk levels in Covid19 context", published in the Multidisciplinary Digital Publishing Institute special issue "Human Activity Recognition Based on Image Sensors and Deep Learning". The repository includes the complete dataset used for the training, validation and testing tasks, in order to detect the presence ou absence of mask and glasses on people, and detect the caruncle zone of people's eyes (both on a thermografic context). For the detection of mask and glasses, there are two different folders: images and labels, each divided in three different subdatasets (train, valid, test). For each image, there is a text document with the exactly same name, where is present the information about each object (in this case, people's faces). This labels information uses the class associated to the object (0: Face_Mask_Eyes, 1: Face_Mask_NoEyes, 2: Face_NoMask_Eyes, 3: Face_NoMask_NoEyes, where Eyes its related to person without glasses and NoEyes related to person with glasses), and the correspondent normalized values of the bounding box of the face (x_center, y_center, width, height). For the detection of the caruncle zone, the image folder has the same format as the previous task, while the labels are shown in JSON format, one for each subdataset, where are presented the bounding box of the face and the pair of coordinates (x, y) of both caruncles.本仓库收录了发表于多学科数字出版机构（Multidisciplinary Digital Publishing Institute）特刊"基于图像传感器与深度学习的人类活动识别"中的论文《新冠疫情背景下基于AI的不同风险等级监测》所述的数据集之一。该仓库包含用于训练、验证与测试任务的完整数据集，可实现两项检测目标：一是检测人体是否佩戴口罩与眼镜，二是在热成像（thermografic）场景下检测人眼的泪阜区（caruncle zone）。针对口罩与眼镜佩戴状态检测任务，仓库设有两个独立文件夹：图像文件夹与标签文件夹，二者均划分为训练、验证、测试三个子数据集。每张图像对应一个同名文本文档，文档中存储了对应目标（此处为人脸）的标注信息。该标签信息包含两类内容：一是与目标绑定的类别标签（0：佩戴口罩且未戴眼镜的人脸，1：佩戴口罩且戴眼镜的人脸，2：未佩戴口罩且未戴眼镜的人脸，3：未佩戴口罩且戴眼镜的人脸；其中"未戴眼镜"对应原文的Eyes，"戴眼镜"对应NoEyes），二是人脸边界框（bounding box）的归一化坐标值（x_center, y_center, width, height）。针对泪阜区检测任务，图像文件夹的格式与前述口罩与眼镜佩戴状态检测任务完全一致，而标签则采用JSON格式，每个子数据集对应一个独立标签文件，其中存储了人脸边界框与两个泪阜的坐标对(x, y)。