US 11,455,831 B2Granted

Method and apparatus for face classification

AssigneeArcSoft, Inc.

InventorsHan Qiu, Sanyong Fang, Kangning Song

Life of this patent

Estimating the term…

Loading patent…

Abstract

The present disclosure provides a method and apparatus for facial classification, which is applied to the field of image processing. The method includes acquiring a color image of a target face, where the color image includes information of at least one channel, inputting the information into a neural network, and classifying, by the neural network, the target face according to the information and a first parameter. The first parameter includes at least one facial category and first parameter data for identifying a facial category of the target face. The method and device of the present disclosure can analyze facial attributes from the content of a facial image, and automatically classify many facial images according to their facial attributes. This reduces the burden of manual classification, thereby allowing facial images to be stored in a clear and orderly fashion and improving classification efficiency.

Claims28

Dependent claims

A method for face classification, comprising:

acquiring a color image of a target face, the color image comprising information of a plurality of channels;
determining that the color image of the target face qualifies for inputting into a neural network in accordance with at least one of the following parameters:
a pitch angle, a drift angle, a roll angle, and a dimension and light conditions;
inputting the information of the plurality of channels of the color image of the target face into the neural network;
and
classifying, by the neural network, the target face in accordance with the information of the plurality of channels of the color image of the target face and a first parameter, the first parameter comprising at least one facial category and data about the first parameter for identifying a facial category of the target face,
wherein, before inputting the information of the plurality of channels of the color image of the target face into the neural network, the method further comprises performing an identical first processing on the information of the plurality of channels of the color image of the target face, the first processing comprising:
normalizing the information of each of the plurality of channels of the color image of the target face with respect to a pixel value of the color image from [0,255] to [0,1],
wherein, before inputting the information of the plurality of channels of the color image of the target face into the neural network, the method further comprises:
obtaining a multidimensional output vector;
and
inputting the multidimensional output vector to layers of the neural network to successively obtain output results pertaining to the target face,
wherein, when the color image of the target face is determined to be not qualified, the method further comprises stopping inputting the color image into the neural network and stopping classification of the color image of the target face, and
wherein the facial category of the target face included in the first parameter comprises gender, race, age range, and age value.

11 more elements

An apparatus for face classification, comprising:

a processor and a neural network, wherein the processor
acquires a color image of a target face, the color image comprising information of a plurality of channels,
determines that the color image of the target face qualifies for inputting into the neural network in accordance with at least one of the following parameters:
a pitch angle, a drift angle, a roll angle, and a dimension and light conditions, and
inputs the information of the plurality of channels of the color image of the target face into the neural network,
wherein the neural network is configured to classify the target face in accordance with the information of the plurality of channels of the color image of the target face and a first parameter, the first parameter comprising at least one facial category and data about the first parameter for identifying a facial category of the target face,
wherein the processor comprises a first processing module that performs, before inputting the information of the plurality of channels of the color image of the target face into the neural network, an identical first processing on the information of the plurality of channels of the color image of the target face,
wherein the first processing module comprises a first normalizing sub-module that normalizes the information of each of the plurality of channels of the color image of the target face with respect to a pixel value from [0,255] to [0,1],
wherein, before inputting the information of the plurality of channels of the color image of the target face into the neural network, the processor further:
obtains a multidimensional output vector;
and
inputs the multidimensional output vector to layers of the neural network to successively obtain output results pertaining to the target face,
wherein, when the color image of the target face is determined to be not qualified, the processor further stops inputting the color image into the neural network and stops classification of the color image of the target face, and
wherein the facial category of the target face included in the first parameter comprises gender, race, age range, and age value.

11 more elements

Dependencies are inferred from the claim text.

Facts

Application: 16/045,204Filed 25 Jul 2018
Priority: CN 201710614131.225 Jul 2017
Published: US 2019/0034704 A131 Jan 2019
Granted: 27 Sep 2022
Term ends: Not estimatedThe source records don’t support an estimate.

Classification15 codes

G06N 3/084
Learning methods: backpropagation, e.g. using gradient descent
G06F 18/2413
Relating to the classification model, e.g. parametric or non-parametric approaches: based on distances to training or reference patterns
G06F 18/24143
Distances to prototypes: distances to neighbourhood prototypes, e.g. restricted Coulomb energy networks [RCEN]
G06N 3/045
Architecture, e.g. interconnection topology: combinations of networks