MaskedFace-Net

该数据集分为正确佩戴口罩人脸数据集（CMFD）和错误佩戴口罩人脸数据集（IMFD），并进一步细分为四个类别：CMFD、IMFD鼻部、IMFD下巴以及IMFD鼻部和嘴部。值得注意的是，该数据集中的样本数量在各个类别之间存在高度不平衡。为了解决这个问题，我们采用了随机抽样和增强技术，最终得到了一个平衡的数据集。该平衡数据集的总规模为110K的训练和验证样本以及28K的测试样本，共计133,783个样本。该数据集的任务是对人脸佩戴口罩的情况及其位置进行分类。

This dataset is divided into two subsets: the Correct Mask-Wearing Face Dataset (CMFD) and the Incorrect Mask-Wearing Face Dataset (IMFD), which are further categorized into four classes: CMFD, IMFD (nose exposed), IMFD (chin exposed), and IMFD (both nose and mouth exposed). Notably, the original dataset exhibits severe class imbalance across all categories. To address this issue, we adopted random sampling and data augmentation techniques to obtain a balanced final dataset. The balanced dataset has a total of 133,783 samples, comprising 110,000 training and validation samples and 28,000 test samples. The task of this dataset is to classify both the correct mask-wearing status of human faces and the specific exposed positions when masks are worn incorrectly.