P3M-10K

来自于公开数据集P3M-10K，含有10,421张保护了人脸隐私的训练集和两个没有人像隐私问题的测试集，经过深度处理，将24深度转为8深度，像素值为0的为前景，1的为人像。能够使用在人像分割领域。 开源github地址：https://github.com/JizhiziLi/P3MDerived from the public dataset P3M-10K, this dataset contains 10,421 training images with privacy-protected human faces, as well as two test sets free of any human-related privacy issues. After deep image processing, 24-bit depth images are converted to 8-bit depth, where pixel value 0 represents the foreground and pixel value 1 corresponds to the human portrait. It can be applied in the field of portrait segmentation. Its open-source GitHub repository is available at https://github.com/JizhiziLi/P3M