568 People - Face Detection & Face 106 Landmarks & Human Body Segmentation Annotation Data in Online Conference Scenes

ライセンス認証を経た製品データセットが、AIプロジェクトのスピーディーな立ち上げをアシストします。

568 People - Face Detection & Face 106 Landmarks & Human Body Segmentation Annotation Data in Online Conference Scenes. The ethnic groups include East Asians, Caucasians, Blacks, and Browns, with a primary focus on young adults. Various indoor office scenes were captured, including meeting rooms, cafes, libraries, and bedrooms. In terms of annotation, each individual consists of 61 to 64 photos, with annotations for detected facial bounding boxes and 106 facial landmarks, as well as segmentation annotations for the human body. The data can be used for tasks such as facial detection, 106 facial landmark recognition, and human body segmentation.

お問い合わせサンプルを入手する

データ仕様

Data size: 568 people, each person contains 61-64 images

Race distribution: 142 Asians, 142 Caucasians, 142 Blacks, and 142 Browns

Nationality distribution: Vietnam, India, Indonesia, Thailand, Pakistan, Russia, France, Germany, etc.

Gender distribution: 323 males, 245 females

Age distribution: ranging from adolescents to the elderly, primarily focusing on young adults

Collection environment: indoor office settings, such as meeting rooms, cafes, libraries, bedrooms, etc.

Collection diversity: various facial angles, facial expressions, age groups, and scenarios

Collection equipment: cellphone, simulating laptop camera perspectives in conference settings

Data format: image data format is .jpg, annotation document format is .json

Annotation content: annotations include detected facial bounding boxes and 106 facial landmarks, as well as body segmentation annotations

Accuracy rate: Face Detection Bounding Box Accuracy: the detection box is considered qualified if the offset does not exceed 5 pixels on all four sides, with a qualification rate of no less than 95%; Facial Landmark Accuracy: the accuracy rate of facial landmark annotation must be no less than 95%; Human Segmentation Accuracy: the mask edge error in the x and y directions must be within 5 pixels to be considered correctly annotated; with the human body as the unit, the annotation accuracy rate should be abov

サンプル紹介

収集対象者からの明確に許可を得た、高品質の製品トレーニングデータセットはが、AIプロジェクトのスピーディーな立ち上げをアシストします。

さっそく始めてみる