Verifying a target object based on confidence coefficients generated by trained models
Embodiments include a method, an electronic device, and a computer program product for information processing. In an example embodiment, a method for information processing includes: acquiring, at a first device, a first feature associated with a target object; applying the first feature to a trained first model deployed at the first device to determine a first confidence coefficient, the first confidence coefficient being associated with probabilities that the first model determines the target object as a real object and as a false object; if the first confidence coefficient is lower than a first threshold confidence coefficient, sending a request for verifying the target object to a second device, the second device being deployed with a trained second model for verifying the target object, and the second model being more complex than the first model; and updating the first model based on a response to the request.
1. A method for information processing, comprising:
acquiring, at a first device, a first feature associated with a target object;
applying the first feature to a trained first model deployed at the first device to determine a first confidence coefficient, the first confidence coefficient being associated with probabilities that the first model determines the target object as a real object and as a false object;
if the first confidence coefficient is lower than a first threshold confidence coefficient, sending a request for verifying the target object to a second device, the second device being deployed with a trained second model for verifying the target object based on a second feature different from any feature applied to the trained first model, the second model being more complex than the first model; and
updating the first model based on a response to the request, the response being from the second device.
2. The method according to claim 1 , wherein the first feature comprises at least one of the following: keystroke mode, gesture, fingerprint, voice, and face.
3. The method according to claim 1 , wherein the response comprises a verification result for the target object, and updating the first model comprises:
retraining the first model based at least on the first feature and the verification result.
4. The method according to claim 3 , wherein retraining the first model comprises:
acquiring a training feature, the training feature being associated with a training object for training the first model;
acquiring a training label, the training label indicating whether the training object is a real object; and
retraining the first model based on the training feature, the training label, the first feature, and the verification result.
5. The method according to claim 1 , wherein the response comprises the first model retrained by the second device, and updating the first model comprises:
replacing the first model with the retrained first model.
6. A non-transitory computer-readable medium with a computer program product tangibly stored thereon and comprising machine-executable instructions, wherein the machine-executable instructions, when executed by a machine, cause the machine to perform the steps of the method according to claim 1 .
7. A method for information processing, comprising:
responsive to a first confidence coefficient being lower than a first threshold confidence coefficient, the first confidence coefficient being associated with probabilities that a trained first model deployed at a first device determines a target object as a real object and as a false object based on a first feature associated with the target object, receiving, at a second device from the first device, a request for verifying the target object;
acquiring a second feature associated with the target object based on the request, the second feature being different from any feature applied to the trained first model;
applying the second feature to a trained second model deployed at the second device to determine a second confidence coefficient, the second confidence coefficient being associated with probabilities that the second model determines the target object as the real object and as the false object; and
sending a response to the request to the first device based on the second confidence coefficient, so as to update the first model deployed at the first device and for verifying the target object, the second model being more complex than the first model.
8. The method according to claim 7 , wherein the second feature comprises at least one of the following: keystroke mode, gesture, fingerprint, voice, and face.
9. The method according to claim 7 , wherein sending the response to the first device comprises:
if the second confidence coefficient is lower than a second threshold confidence coefficient, acquiring a third feature associated with the target object;
applying the third feature to a trained third model deployed at the second device to determine a third confidence coefficient, the third confidence coefficient being associated with probabilities that the third model classifies the target object into the real object and into the false object, and the third model being more complex than the second model; and
sending the response to the first device based on the third confidence coefficient.
10. The method according to claim 7 , wherein sending the response to the first device comprises:
if the second confidence coefficient exceeds a second threshold confidence coefficient, generating a verification result that the target object is the real object;
including the verification result in the response; and
sending the response to the first device.
11. The method according to claim 7 , wherein sending the response to the first device comprises:
if the second confidence coefficient exceeds a second threshold confidence coefficient, generating a verification result that the target object is the real object;
retraining the first model based at least on the first feature and the verification result; and
sending the response to the first device, the response comprising the retrained first model.
12. The method according to claim 11 , wherein retraining the first model comprises:
acquiring a training feature, the training feature being associated with a training object for training the first model;
acquiring a training label, the training label indicating whether the training object is a real object; and
retraining the first model based on the training feature, the training label, the first feature, and the verification result.
13. The method according to claim 11 , wherein sending the response comprises:
adjusting parameters of layers in the retrained first model to compress the retrained first model;
including the compressed first model in the response; and
sending the response to the first device.
14. A non-transitory computer-readable medium with a computer program product tangibly stored thereon and comprising machine-executable instructions, wherein the machine-executable instructions, when executed by a machine, cause the machine to perform the steps of the method according to claim 7 .
15. An apparatus, comprising:
at least one processing unit; and
at least one memory that is coupled to the at least one processing unit and stores instructions for execution by the at least one processing unit, wherein the instructions, when executed by the at least one processing unit, cause the apparatus to perform actions comprising:
acquiring, at a first device, a first feature associated with a target object;
applying the first feature to a trained first model deployed at the first device to determine a first confidence coefficient, the first confidence coefficient being associated with probabilities that the first model determines the target object as a real object and as a false object;
if the first confidence coefficient is lower than a first threshold confidence coefficient, sending a request for verifying the target object to a second device, the second device being deployed with a trained second model for verifying the target object based on a second feature different from any feature applied to the trained first model, the second model being more complex than the first model; and
updating the first model based on a response to the request, the response being from the second device.
16. The apparatus according to claim 15 , wherein the first feature comprises at least one of the following: keystroke mode, gesture, fingerprint, voice, and face.
17. The apparatus according to claim 15 , wherein the response comprises a verification result for the target object, and updating the first model comprises:
retraining the first model based at least on the first feature and the verification result.
18. The apparatus according to claim 17 , wherein retraining the first model comprises:
acquiring a training feature, the training feature being associated with a training object for training the first model;
acquiring a training label, the training label indicating whether the training object is a real object; and
retraining the first model based on the training feature, the training label, the first feature, and the verification result.
19. The apparatus according to claim 15 , wherein the response comprises the first model retrained by the second device, and updating the first model comprises:
replacing the first model with the retrained first model.
20. The apparatus according to claim 15 wherein the at least one processing unit and the at least one memory implement the first device and the second device, wherein the instructions, when executed by the at least one processing unit, cause the second device to perform actions comprising:
receiving the request for verifying the target object, from the first device;
acquiring the second feature associated with the target object based on the request;
applying the second feature to the trained second model deployed at the second device to determine a second confidence coefficient, the second confidence coefficient being associated with probabilities that the second model determines the target object as the real object and as the false object; and
sending the response to the request to the first device based on the second confidence coefficient, so as to update the first model deployed at the first device and for verifying the target object.