Human face image classification method and apparatus, and server

1. A human face image classification method, comprising following steps:

acquiring a human face image to be classified;

inputting the human face image into a pre-set convolutional neural network model, and identifying gender information of the human face image according to intermediate data output by a convolutional layer of the convolutional neural network model; and

carrying out pre-set content understanding classification on the human face image in a range defined by the gender information according to final data output by the convolutional layer of the convolutional neural network model, so that data for deciding a classification result output by the convolutional neural network model comprises difference attribute data for distinguishing different genders;

wherein the convolutional neural network model comprises: a first classification set and a second classification set;

the first classification set identifies the gender information of the human face image according to the intermediate data output by the convolutional layer of the convolutional neural network model; and

the second classification set carries out the pre-set content understanding classification on the human face image in the range defined by the gender information according to the final data output by the convolutional layer of the convolutional neural network model.

2. The human face image classification method according to claim 1 , wherein the first classification set at least comprises two fully connected layers and one classification layer; the first classification set takes intermediate data output by a penultimate convolutional layer in convolutional layers as an input signal; the second classification set comprises at least two fully connected layers and one classification layer; and the second classification set takes final data output by a last convolutional layer in the convolutional layers as an input signal.

3. The human face image classification method according to claim 1 , wherein the convolutional neural network model is formed by training through following steps:

acquiring a training sample set marked with gender information and desired classification information;

inputting the training sample set into the convolutional neural network model to train the first classification set until the first classification set is convergent;

inputting the training sample set into the convolutional neural network model after the first classification set is convergent, training the second classification set until the second classification set is convergent; and

inputting the training sample set into the convolutional neural network model after the second classification set is convergent, and finely adjusting the first classification set and the second classification set by using a pre-set learning rate until the convolutional neural network model is in a convergent state.

4. The human face image classification method according to claim 3 , wherein inputting the training sample set into the convolutional neural network model to train the first classification set until the first classification set is convergent comprises following steps:

inputting the training sample set into the convolutional neural network model, and acquiring a gender classification result output by the first classification set;

comparing whether the marked gender information is consistent with the gender classification result by using a loss stop function;

repeatedly, cyclically and iteratively updating weights of convolutional layers in front of the first classification set until the marked gender information is consistent with the gender classification result when the marked gender information is inconsistent with the gender classification result; and

maintaining the weights of the convolutional layers in front of the first classification set when the marked gender information is consistent with the gender classification result.

5. The human face image classification method according to claim 4 , wherein inputting the training sample set into the convolutional neural network model after the first classification set is convergent, and training the second classification set until the second classification set is convergent comprise following steps:

inputting the training sample set into the convolutional neural network model, and acquiring an excitation classification result output by the second classification set;

comparing whether the desired classification information is consistent with the excitation classification result by using the loss stop function;

repeatedly, cyclically and iteratively updating weights of convolutional layers between the first classification set and the second classification set until the desired classification information is consistent with the excitation classification result when the desired classification information is inconsistent with the excitation classification result; and

maintaining the weights of the convolutional layers between the first classification set and the second classification set when the marked gender information is consistent with the gender classification result.

6. The human face image classification method according to claim 1 , wherein after inputting the human face image into the pre-set convolutional neural network model, and identifying the gender information of the human face image according to the intermediate data output by the convolutional layer of the convolutional neural network model, the method further comprises following steps:

grading a beauty score of the human face image in the range defined by the gender information according to the final data output by the convolutional layer of the convolutional neural network model, so that data for deciding a beauty score output by the convolutional neural network model comprises difference attribute data for distinguishing different genders.

7. The human face image classification method according to claim 1 , wherein

the pre-set content understanding classification comprises: image similarity comparison, race classification identification, or age identification.

8. A non-transitory computer readable storage medium, configured to store a computer program, wherein the computer program is executed to implement: the human face image classification method according to claim 1 .

9. A non-transitory computer program product, wherein the computer program product is configured to, when executed, implement: the human face image classification method according to claim 1 .

10. The server according to claim 9 , wherein the first classification set at least comprises two fully connected layers and one classification layer; the first classification set takes intermediate data output by a penultimate convolutional layer in convolutional layers as an input signal; the second classification set comprises at least two fully connected layers and one classification layer; and the second classification set takes final data output by a last convolutional layer in the convolutional layers as an input signal.

11. The server according to claim 9 , wherein the one or more application programs are further configured to form the convolutional neural network model by training through following steps:

acquiring a training sample set marked with gender information and desired classification information;

inputting the training sample set into the convolutional neural network model to train the first classification set until the first classification set is convergent;

inputting the training sample set into the convolutional neural network model after the first classification set is convergent, training the second classification set until the second classification set is convergent; and

inputting the training sample set into the convolutional neural network model after the second classification set is convergent, and finely adjusting the first classification set and the second classification set by using a pre-set learning rate until the convolutional neural network model is in a convergent state.

12. The server according to claim 11 , inputting the training sample set into the convolutional neural network model to train the first classification set until the first classification set is convergent comprises following steps:

inputting the training sample set into the convolutional neural network model, and acquiring a gender classification result output by the first classification set;

comparing whether the marked gender information is consistent with the gender classification result by using a loss stop function;

repeatedly, cyclically and iteratively updating weights of convolutional layers in front of the first classification set until the marked gender information is consistent with the gender classification result when the marked gender information is inconsistent with the gender classification result; and

maintaining the weights of the convolutional layers in front of the first classification set when the marked gender information is consistent with the gender classification result.

13. The server according to claim 12 , wherein inputting the training sample set into the convolutional neural network model after the first classification set is convergent, and training the second classification set until the second classification set is convergent comprise following steps:

inputting the training sample set into the convolutional neural network model, and acquiring an excitation classification result output by the second classification set;

comparing whether the desired classification information is consistent with the excitation classification result by using the loss stop function;

repeatedly, cyclically and iteratively updating weights of convolutional layers between the first classification set and the second classification set until the desired classification information is consistent with the excitation classification result when the desired classification information is inconsistent with the excitation classification result; and

maintaining the weights of the convolutional layers between the first classification set and the second classification set when the marked gender information is consistent with the gender classification result.

14. A server, comprising:

one or more processors;

a memory;

one or more application programs;

wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors; and the one or more application programs are configured to execute a human face image classification method comprising:

acquiring a human face image to be classified;

inputting the human face image into a pre-set convolutional neural network model, and identifying gender information of the human face image according to intermediate data output by a convolutional layer of the convolutional neural network model; and

carrying out pre-set content understanding classification on the human face image in a range defined by the gender information according to final data output by the convolutional layer of the convolutional neural network model, so that data for deciding a classification result output by the convolutional neural network model comprises difference attribute data for distinguishing different genders;

wherein the convolutional neural network model comprises: a first classification set and a second classification set;

the first classification set identifies the gender information of the human face image according to the intermediate data output by the convolutional layer of the convolutional neural network model; and

the second classification set carries out the pre-set content understanding classification on the human face image in the range defined by the gender information according to the final data output by the convolutional layer of the convolutional neural network model.

15. The server according to claim 14 , wherein after inputting the human face image into the pre-set convolutional neural network model, and identifying the gender information of the human face image according to the intermediate data output by the convolutional layer of the convolutional neural network model, the one or more application programs are further configured to execute the human face image classification method comprising following steps:

grading a beauty score of the human face image in the range defined by the gender information according to the final data output by the convolutional layer of the convolutional neural network model, so that data for deciding a beauty score output by the convolutional neural network model comprises difference attribute data for distinguishing different genders.

16. The server according to claim 14 , the pre-set content understanding classification comprises: image similarity comparison, race classification identification, or age identification.