Method, electronic device, and computer program product for target image processing

Patent metadata

Field	Value
Publication number	US-12561959-B2
Application number	US-202217988484-A
Country	US
Kind code	B2
Filing date	Nov 16, 2022
Priority date	Oct 21, 2022
Publication date	Feb 24, 2026
Grant date	Feb 24, 2026

Abstract

Embodiments disclosed herein include a method, an electronic device, and a computer program product for target image processing. The method includes receiving a target image and generating a first Shapley value for a feature of the target image based on the received target image. The method further includes sending, in response to satisfying a predetermined condition, a request for acquiring a second Shapley value to a cloud server. The method further includes receiving the second Shapley value for a latent feature of the target image from the cloud server, where the second Shapley value is more accurate than the first Shapley value. In some embodiments, through joint collaboration between a terminal device such as an edge device and a cloud server, rapid calculation of a Shapley value can be achieved at the terminal device, and accurate calculation of a Shapley value can be achieved at the cloud server.

First claim

What is claimed is:

1 . A method, comprising:
receiving a target image; generating a first Shapley value for a feature of the target image based on the received target image; sending, in response to satisfying a predetermined condition, a request for acquiring a second Shapley value to a cloud server; and receiving the second Shapley value for a latent feature of the target image from the cloud server; wherein the second Shapley value is more accurate than the first Shapley value; wherein the first Shapley value is generated by an approximation model; and wherein the approximation model is trained based on a uniform distribution and a feature distribution of the feature.

2 . The method according to claim 1 , further comprising:
encoding the target image by using an encoder, to acquire the feature of the target image; acquiring the latent feature in the feature; and sending the latent feature to the cloud server, so that the cloud server generates the second Shapley value based on the latent feature.

3 . The method according to claim 2 , wherein the encoder is a part of an auto-encoder, and the auto-encoder further comprises a decoder deployed in the cloud server.

4 . The method according to claim 1 , wherein the method further comprises:
fine-tuning the approximation model based on the first Shapley value and the second Shapley value.

5 . The method according to claim 1 , wherein the method further comprises:
training the approximation model at least in part by: acquiring a value function of a reference model associated with the feature; determining a value function expectation function based on the value function under the uniform distribution and the feature distribution of the feature; and minimizing the value function expectation function to train the approximation model.

6 . The method according to claim 5 , wherein the reference model receives the feature and outputs probability distribution with respect to a true value according to the feature.

7 . The method according to claim 5 , further comprising:
training the reference model at least in part by: acquiring a masked feature under the feature distribution of the feature; determining a distribution expectation function based on the masked feature; and minimizing the distribution expectation function to train the reference model.

8 . An electronic device, comprising:
at least one processor; and at least one memory coupled to the at least one processor and having instructions stored therein, wherein the instructions, when executed by the at least one processor, cause the electronic device to perform actions comprising: receiving a target image; generating a first Shapley value for a feature of the target image based on the received target image; sending, in response to satisfying a predetermined condition, a request for acquiring a second Shapley value to a cloud server; and receiving the second Shapley value for a latent feature of the target image from the cloud server; wherein the second Shapley value is more accurate than the first Shapley value; wherein the first Shapley value is generated by an approximation model; and wherein the approximation model is trained based on a uniform distribution and a feature distribution of the feature.

9 . The electronic device according to claim 8 , wherein the instructions, when executed by the at least one processor, further cause the electronic device to perform:
encoding the target image by using an encoder, to acquire the feature of the target image; acquiring the latent feature in the feature; and sending the latent feature to the cloud server, so that the cloud server generates the second Shapley value based on the latent feature.

10 . The electronic device according to claim 9 , wherein the encoder is a part of an auto-encoder, and the auto-encoder further comprises a decoder deployed in the cloud server.

11 . The electronic device according to claim 8 , wherein the instructions, when executed by the at least one processor, further cause the electronic device to perform:
fine-tuning the approximation model based on the first Shapley value and the second Shapley value.

12 . The electronic device according to claim 8 , wherein the instructions, when executed by the at least one processor, further cause the electronic device to train the approximation model at least in part by:
acquiring a value function of a reference model associated with the feature; determining a value function expectation function based on the value function under the uniform distribution and the feature distribution of the feature; and minimizing the value function expectation function to train the approximation model.

13 . The electronic device according to claim 12 , wherein the reference model receives the feature and outputs probability distribution with respect to a true value according to the feature.

14 . The electronic device according to claim 12 , wherein the instructions, when executed by the at least one processor, further cause the electronic device to train the reference model at least in part by:
acquiring a masked feature under the feature distribution of the feature; determining a distribution expectation function based on the masked feature; and minimizing the distribution expectation function to train the reference model.

15 . A computer program product comprising a non-transitory computer-readable medium storing machine-executable instructions that, when executed by a machine, cause the machine to perform actions comprising:
receiving a target image; generating a first Shapley value for a feature of the target image based on the received target image; sending, in response to satisfying a predetermined condition, a request for acquiring a second Shapley value to a cloud server; and receiving the second Shapley value for a latent feature of the target image from the cloud server; wherein the second Shapley value is more accurate than the first Shapley value; wherein the first Shapley value is generated by an approximation model; and wherein the approximation model is trained based on a uniform distribution and a feature distribution of the feature.

16 . The computer program product according to claim 15 , wherein the machine-executable instructions, when executed, further cause the machine to perform actions comprising:
encoding the target image by using an encoder, to acquire the feature of the target image; acquiring the latent feature in the feature; and sending the latent feature to the cloud server, so that the cloud server generates the second Shapley value based on the latent feature.

17 . The computer program product according to claim 16 , wherein the encoder is a part of an auto-encoder, and the auto-encoder further comprises a decoder deployed in the cloud server.

18 . The computer program product according to claim 16 , wherein the machine-executable instructions, when executed, further cause the machine to perform an action comprising:
fine-tuning the approximation model based on the first Shapley value and the second Shapley value.

Assignees

• Dell Products Lp

Inventors

• Wang Zijia
• Li Sanping
• Ni Jiacheng
• Jia Zhen

Classifications

• G06V10/763 Primary

  Non-hierarchical techniques, e.g. based on statistics of modelling distributions · CPC title

• G06V10/82 Primary

  using neural networks · CPC title

• G06V10/776 Primary

  Validation; Performance evaluation · CPC title

• G06N20/00

  Machine learning · CPC title