Apple Inc.

Atulit Kumar

Thorsten Gernoth

Haitao Guo

Ian R Fasel

Feng Tang

Onur Canturk Hamsici

Sofien Bouaziz

Thibaut Weise

Joerg A Liebelt

Sarah Amsellem

An image captured using a camera on a device (e.g., a mobile device) may be operated on by one or more processes to determine properties of a user's face in the image. A first process may determine one or more first properties of the user's face in the image. A second process operating downstream from the first process may determine at least one second property of the user's face in the image. The second process may use at least one of the first properties from the first process to determine the second property.

Network performance by including attributes

A neural network may implement a face detection process on a mobile device or computer system. An image captured using a camera on the device may be processed using the face detection process. The face detection process may provide a bounding box for the face detected in the image. The face detected in the image may have any orientation. The face detected in the image may include either a whole face of the user or only a partial face of the user. The face detection process may provide face detection that is substantially invariant to exposure in the captured image. The face may also be detected over a wide range of distances between the camera and the face of the user.

Robust face detection

A single network encodes and decodes an image captured using a camera on a device. The single network detects if a face is in the image. If a face is detected in the image, the single network determines properties of the face in the image and outputs the properties along with the face detection output. Properties of the face may be determined by sharing the task for face detection. Properties of the face that are output along with the face detection output include the location of the face, the pose of the face, and/or the distance of the face from the camera.

Face detection, pose estimation, and distance from a camera estimation using a single network

An estimate of distance between a user and a camera on a device is used to determine an illumination pattern density used for speckle pattern illumination of the user in subsequent images. The distance may be estimated using an image captured when the user is illuminated with flood infrared illumination. Either a sparse speckle (dot) pattern illumination pattern or a dense speckle pattern illumination pattern is used depending on the distance between the user's face and the camera.

Determining sparse versus dense pattern illumination

Techniques are disclosed for creating avatars from image data of a person. According to these techniques, spatial facial attributes of a subject may be measured from an image representing the subject. The measured facial attributes may be matching to a three-dimensional avatar template. Other attributes of the subject may be identified, such as hair type, hair color, eye color, skin color and the like. For hair type, hair may be generated for the avatar by measuring spatial locations of hair of the subject and the measured hair locations may be compared to locations of hair represented by a plurality of hair templates. A matching hair template may be selected from the plurality of hair templates, which may be used in generating the avatar. An avatar may be generated from the matching avatar template, which may be deformed according to the measured spatial facial attributes, and from the other attributes.

Growing community of inventors

Cupertino, CA, United States of America

Atulit Kumar