70,846 Images – Human Face Segmentation Data
Datatang has passed the certification of ISO27001 Information Security Management System and ISO9001 Quality Management System.
70,846 Images – Human Face Segmentation Data. The data includes pure color background, indoor scenes and outdoor scenes. The data covers female peole and male people. The race distribution includes yellow race, black race and white race. The age distribution ranges from baby to senior. The data contains simple face poses and complicated expressions(large-angle tilt of face, closing eye, glower, pucker, opening mouth, etc.). In terms of annotation, we adpoted pixel-wise segmentation annotations on human face, the five sense organs, body and appendages. The data can be used for tasks such as face segmentation.
- Data size
- 70,846 images, there is only one face in an image
- Population distribution
- race distribution: 32,235 faces of yellow race, 29,501 faces of white race, 9,110 faces of black race; gender distribution: 39,917 male faces and 30,929 female faces; age distribution: baby, teenager, young, midlife and senior
- Collection environment
- including pure color background, indoor scenes and outdoor scenes
- Data diversity
- multiple scenes, multiple ages, multiple races, complicated expressions (closing eye, glower, pucker, opening mouth, etc.), and multiple appendages
- Image Parameter
- Data format: the image data is in .jpg or .png format, the annotation file is in .json or .psd format; the human face resolution is not lower than 128*128, and pupillary distance is not less than 60 pixels
- Annotation content
- segmentation annotation of human face, the five sense organs, body and appendages
- the mask edge location errors in x and y directions are less than 3 pixels, which is considered as a qualified annotation; the annotation part (id) is regarded as the unit, the accuracy rate of segmentation annotation shall be more than 97%