Method, device, and computer program product for model updating
The present disclosure relates to a method, a device, and a computer program product for model updating. The method includes: acquiring a first image set and first annotation information, wherein the first annotation information indicates whether a corresponding image in the first image set includes a target object; updating a first version of an object verification model using the first image set and the first annotation information to obtain a second version, wherein the first version of the object verification model has been deployed to determine whether an input image includes the target object; determining the accuracy of the second version of the object verification model; and updating, if it is determined that the accuracy is lower than a preset accuracy threshold, the second version of the object verification model using a second image set and second annotation information to obtain a third version of the object verification model.
1. A model updating method, comprising:
acquiring a first image set and first annotation information, wherein the first annotation information indicates whether a corresponding image in the first image set includes a target object;
updating a first version of an object verification model using the first image set and the first annotation information to obtain a second version of the object verification model, wherein the first version of the object verification model has been deployed to determine whether an input image includes the target object;
determining the accuracy of the second version of the object verification model; and
updating, responsive to it being determined that the accuracy is lower than a preset accuracy threshold, the second version of the object verification model using a second image set and second annotation information to obtain a third version of the object verification model, wherein the number of images in the second image set is greater than the number of images in the first image set, and the second annotation information indicates whether a corresponding image in the second image set includes a desired object.
2. The method according to claim 1 , wherein acquiring the first image set and the first annotation information comprises:
acquiring at least one input image collected in the use of the first version of the object verification model and a verification result corresponding to the at least one image, wherein the verification result indicates whether the at least one image includes the target object;
including the at least one input image into the first image set; and
determining the first annotation information based on the verification result.
3. The method according to claim 2 , wherein acquiring the at least one input image comprises acquiring at least one of the following:
a first input image that is determined by the object verification model to include the target object, wherein the first annotation information is determined to indicate that the first input image includes the target object,
a second input image that is determined by the object verification model to not include the target object, wherein the verification result of the object verification model is used to determine a permission, an object in the second input image is determined to have the permission, and the first annotation information is determined to indicate that the second input image includes the target object, and
a third input image that is determined by the object verification model to not include the target object, wherein an object in the third input image is determined to not have the permission, and the first annotation information is determined to indicate that the second input image does not include the target object,
a reference image that is provided for verification by the object verification model, wherein the first annotation information is determined to indicate that the reference image includes the target object.
4. The method according to claim 1 , wherein updating the first version of the object verification model comprises:
training the first version of the object verification model using the first image set and the first annotation information to meet a first convergence condition of a first objective function; and
wherein updating the second version of the object verification model comprises:
training the second version of the object verification model using the second image set and the second annotation information to meet a second convergence condition of a second objective function, wherein the complexity of meeting the second convergence condition is greater than the complexity of meeting the first convergence condition.
5. The method according to claim 1 , wherein updating the first version of the object verification model comprises:
updating the first version of the object verification model using training that is based on mini-batch stochastic gradient descent.
6. The method according to claim 1 , wherein updating the first version of the object verification model comprises:
training the first version of the object verification model based on a preset learning rate using the first image set and the first annotation information, wherein the preset learning rate is less than a preset threshold.
7. The method according to claim 1 , wherein the target object comprises a target face.
8. An electronic device, comprising:
at least one processor; and
at least one memory storing computer program instructions, wherein the at least one memory and the computer program instructions are configured to cause, together with the at least one processor, the electronic device to perform actions including:
acquiring a first image set and first annotation information, wherein the first annotation information indicates whether a corresponding image in the first image set includes a target object;
updating a first version of an object verification model using the first image set and the first annotation information to obtain a second version of the object verification model, wherein the first version of the object verification model has been deployed to determine whether an input image includes the target object;
determining the accuracy of the second version of the object verification model; and
updating, responsive to it being determined that the accuracy is lower than a preset accuracy threshold, the second version of the object verification model using a second image set and second annotation information to obtain a third version of the object verification model, wherein the number of images in the second image set is greater than the number of images in the first image set, and the second annotation information indicates whether a corresponding image in the second image set includes a desired object.
9. The device according to claim 8 , wherein acquiring the first image set and the first annotation information includes:
acquiring at least one input image collected in the use of the first version of the object verification model and a verification result corresponding to the at least one image, wherein the verification result indicates whether the at least one image includes the target object;
including the at least one input image into the first image set; and
determining the first annotation information based on the verification result.
10. The device according to claim 9 , wherein acquiring the at least one input image comprises acquiring at least one of the following:
a first input image that is determined by the object verification model to include the target object, wherein the first annotation information is determined to indicate that the first input image includes the target object,
a second input image that is determined by the object verification model to not include the target object, wherein the verification result of the object verification model is used to determine a permission, an object in the second input image is determined to have the permission, and the first annotation information is determined to indicate that the second input image includes the target object, and
a third input image that is determined by the object verification model to not include the target object, wherein an object in the third input image is determined to not have the permission, and the first annotation information is determined to indicate that the second input image does not include the target object,
a reference image that is provided for verification by the object verification model, wherein the first annotation information is determined to indicate that the reference image includes the target object.
11. The device according to claim 8 , wherein updating the first version of the object verification model comprises:
training the first version of the object verification model using the first image set and the first annotation information to meet a first convergence condition of a first objective function; and
wherein updating the second version of the object verification model comprises:
training the second version of the object verification model using the second image set and the second annotation information to meet a second convergence condition of a second objective function, wherein the complexity of meeting the second convergence condition is greater than the complexity of meeting the first convergence condition.
12. The device according to claim 8 , wherein updating the first version of the object verification model comprises:
updating the first version of the object verification model using training that is based on mini-batch stochastic gradient descent.
13. The device according to claim 8 , wherein updating the first version of the object verification model comprises:
training the first version of the object verification model based on a preset learning rate using the first image set and the first annotation information, wherein the preset learning rate is less than a preset threshold.
14. The device according to claim 8 , wherein the target object comprises a target face.
15. A computer program product tangibly stored on a non-transitory computer-readable medium and comprising computer-executable instructions, wherein the computer-executable instructions, when executed, cause a device to perform actions comprising:
acquiring a first image set and first annotation information, wherein the first annotation information indicates whether a corresponding image in the first image set includes a target object;
updating a first version of an object verification model using the first image set and the first annotation information to obtain a second version of the object verification model, wherein the first version of the object verification model has been deployed to determine whether an input image includes the target object;
determining the accuracy of the second version of the object verification model; and
updating, responsive to it being determined that the accuracy is lower than a preset accuracy threshold, the second version of the object verification model using a second image set and second annotation information to obtain a third version of the object verification model, wherein the number of images in the second image set is greater than the number of images in the first image set, and the second annotation information indicates whether a corresponding image in the second image set includes a desired object.
16. The computer program product according to claim 15 , wherein acquiring the first image set and the first annotation information comprises:
acquiring at least one input image collected in the use of the first version of the object verification model and a verification result corresponding to the at least one image, wherein the verification result indicates whether the at least one image includes the target object;
including the at least one input image into the first image set; and
determining the first annotation information based on the verification result.
17. The computer program product according to claim 16 , wherein acquiring the at least one input image comprises acquiring at least one of the following:
a first input image that is determined by the object verification model to include the target object, wherein the first annotation information is determined to indicate that the first input image includes the target object,
a second input image that is determined by the object verification model to not include the target object, wherein the verification result of the object verification model is used to determine a permission, an object in the second input image is determined to have the permission, and the first annotation information is determined to indicate that the second input image includes the target object, and
a third input image that is determined by the object verification model to not include the target object, wherein an object in the third input image is determined to not have the permission, and the first annotation information is determined to indicate that the second input image does not include the target object,
a reference image that is provided for verification by the object verification model, wherein the first annotation information is determined to indicate that the reference image includes the target object.
18. The computer program product according to claim 15 , wherein updating the first version of the object verification model comprises:
training the first version of the object verification model using the first image set and the first annotation information to meet a first convergence condition of a first objective function; and
wherein updating the second version of the object verification model comprises:
training the second version of the object verification model using the second image set and the second annotation information to meet a second convergence condition of a second objective function, wherein the complexity of meeting the second convergence condition is greater than the complexity of meeting the first convergence condition.
19. The computer program product according to claim 15 , wherein updating the first version of the object verification model comprises:
updating the first version of the object verification model using training that is based on mini-batch stochastic gradient descent.
20. The computer program product according to claim 15 , wherein updating the first version of the object verification model comprises:
training the first version of the object verification model based on a preset learning rate using the first image set and the first annotation information, wherein the preset learning rate is less than a preset threshold.