CN-121996118-A - Mark processing method and device based on model

Abstract

The embodiment of the specification provides a mark processing method and a device based on a model, wherein the mark processing method based on the model comprises the steps of acquiring multi-mode data submitted by a client and inputting a mark generation model for mark generation in the process of carrying out mark processing based on the model, acquiring mark data containing a plurality of candidate marks, constructing a mark relation diagram based on the mark data, generating an editing interaction network corresponding to the mark relation diagram and returning to the client, further determining mark editing parameters according to the editing instructions under the condition that the editing instructions submitted by the client through the editing interaction network are acquired, calling the mark generation model based on the mark editing parameters for carrying out editing interaction processing corresponding to the editing instructions, and realizing mark generation and mark processing through the mark generation model on the basis of being matched with the client.

Inventors

• ZHOU CHUNXIAN

Assignees

• 支付宝(杭州)数字服务技术有限公司

Dates

Publication Date

20260508

Application Date

20260116

Claims (19)

1. 1. A model-based token processing method, comprising: acquiring multi-mode data submitted by a client, inputting a mark generation model for mark generation, and acquiring mark data containing a plurality of candidate marks; Constructing a mark relation diagram based on the mark data, generating an editing interaction network corresponding to the mark relation diagram and returning to the client; Determining a mark editing parameter according to an editing instruction submitted by the client through the editing interaction network; And calling the mark generation model based on the mark editing parameters to carry out editing interaction processing corresponding to the editing instruction.
2. 2. The model-based logo processing method as claimed in claim 1, wherein the determining logo editing parameters according to editing instructions submitted by the client through the editing interaction network comprises: Generating a mark editing parameter corresponding to the editing node according to the editing node triggered by the editing instruction in the editing interaction network, or generating the mark editing parameter according to the editing node associated with the editing instruction in the editing interaction network and input editing data.
3. 3. The method for processing the mark based on the model according to claim 2, wherein the generating the mark editing parameter corresponding to the editing node by the editing node triggered in the editing interaction network according to the editing instruction comprises: generating a mark generation parameter corresponding to any one of a plurality of generation nodes according to a trigger instruction of the generation node after the source mark object in the editing interaction network is triggered.
4. 4. The model-based logo processing method as claimed in claim 2, the generating the logo editing parameters according to editing nodes associated in the editing interaction network according to the editing instructions and input editing data, comprising: And generating a mark editing parameter of the candidate mark according to the candidate mark triggered by the editing instruction in the editing interaction network and editing data input for the candidate mark.
5. 5. The model-based logo processing method as claimed in claim 4, wherein the editing interaction processing is implemented by the following manner: extracting the mark characteristics and the low-dimensional structural characteristics of the candidate marks, and extracting the editing semantic characteristics of the editing data; And carrying out iterative adjustment on the low-dimensional structural features based on the fusion features of the mark features and the editing semantic features, and obtaining a secondary editing mark after the iterative adjustment is completed.
6. 6. The model-based logo processing method as claimed in claim 1, wherein the determining logo editing parameters according to editing instructions submitted by the client through the editing interaction network comprises: determining associated candidate marks according to displacement data of the displacement editing instruction after the candidate marks in the editing interaction network are selected; generating a sign fusion parameter corresponding to the candidate sign and the associated candidate sign.
7. 7. The model-based token processing method of claim 6, the generating token fusion parameters for both the candidate token and the associated candidate token, comprising: And determining respective attribute weight labels according to the hierarchical relationship and/or the displacement editing relationship of the candidate marks and the associated candidate marks, and generating the mark fusion parameters based on the attribute weight labels and the respective attribute prompt words.
8. 8. The model-based logo processing method as claimed in claim 7, wherein the editing interaction processing is implemented by the following manner: Carrying out weight quantization on the attribute weight labels to obtain attribute weight values, and carrying out weighted fusion on the attribute features of the candidate marks and the associated candidate marks according to the attribute weight values to obtain weighted fusion features; And constructing a cooperative constraint feature based on the weighted fusion feature and the attribute weight value, and performing feature diffusion iteration based on the cooperative constraint feature, the attribute weight value, the candidate mark and the low-dimensional structural feature of each associated candidate mark to obtain a fusion mark.
9. 9. The model-based logo processing method as claimed in claim 1, wherein the logo data comprises attribute prompt words of candidate logos, wherein the attribute prompt words comprise style prompt words, preference prompt words, visual prompt words and/or industry prompt words; And if the attribute node of any candidate mark is triggered, displaying the attribute prompt word of the any candidate mark.
10. 10. The model-based logo processing method as claimed in claim 1, the logo editing parameters comprising logo update data for at least one modality submitted for a source logo object in the editing interaction network; Correspondingly, the editing interaction processing is realized in the following manner: Extracting attribute features and content features of the candidate marks, and extracting adaptation dimension features of the mark update data; performing feature alignment and feature fusion on the attribute features and the content features and the adaptation dimension features to obtain fusion features; And performing flag iteration update based on the fusion features and the low-dimensional structural features of the candidate flags to obtain updated candidate flags.
11. 11. The model-based logo processing method as claimed in claim 1, wherein the determining logo editing parameters according to editing instructions submitted by the client through the editing interaction network comprises: determining candidate marks associated with any one of the generation nodes according to the node coordinates of any one of the generation nodes triggered in the plurality of generation nodes of the source mark object in the editing interaction network and the coordinates of each candidate mark of the editing interaction network; Generating a mark generation parameter according to the coordinate association relation between any generation node and the associated candidate mark.
12. 12. A model-based token processing method, comprising: Acquiring multi-mode data generated by a mark input by a user and submitting the multi-mode data to a server; Receiving and displaying an editing interaction network which is returned by the server and is generated by constructing a sign relation graph based on sign data, wherein the sign data is obtained by inputting the multi-mode data into a sign generation model for sign generation; and acquiring an editing instruction submitted by the user and submitting the editing instruction to the server so as to determine a mark editing parameter according to the editing instruction and call the mark generation model to carry out editing interaction processing.
13. 13. The model-based logo processing method as claimed in claim 12, the editing instructions comprising at least one of: And aiming at the trigger instruction of the editing node in the editing interaction network, aiming at the trigger instruction of any one of a plurality of generating nodes displayed after the source mark object in the editing interaction network is triggered, and aiming at the displacement editing instruction of the selected candidate mark in the editing interaction network.
14. 14. The method for processing the mark based on the model according to claim 12, after the step of receiving and displaying the edited interaction network that is returned by the server and is generated by constructing the mark relation graph based on the mark data is performed, further comprising: And acquiring mark update data of at least one mode submitted by the source mark object in the editing interaction network and submitting the mark update data to the server, or acquiring editing data of candidate mark input triggered by the editing instruction in the editing interaction network and submitting the editing data to the server.
15. 15. A model-based logo processing apparatus comprising: the mark generation module is configured to acquire multi-mode data submitted by a client and input a mark generation model to generate marks so as to acquire mark data containing a plurality of candidate marks; the relation diagram construction module is configured to construct a mark relation diagram based on the mark data, generate an editing interaction network corresponding to the mark relation diagram and return to the client; the parameter determining module is configured to determine mark editing parameters according to editing instructions submitted by the client through the editing interaction network; And the editing processing module is configured to call the mark generation model based on the mark editing parameters to carry out editing interaction processing corresponding to the editing instruction.
16. 16. A model-based logo processing apparatus comprising: the data submitting module is configured to acquire multi-mode data generated by the progress mark input by a user and submit the multi-mode data to the server; The interactive network display module is configured to receive and display an editing interactive network which is returned by the server and is generated by constructing a sign relation graph based on sign data, wherein the sign data is obtained by inputting the multi-mode data into a sign generation model for sign generation; the instruction submitting module is configured to acquire an editing instruction submitted by the user and submit the editing instruction to the server so as to determine a mark editing parameter according to the editing instruction and call the mark generation model to carry out editing interaction processing.
17. 17. A model-based logo processing apparatus comprising: And a memory configured to store computer-executable instructions that, when executed, cause the processor to: acquiring multi-mode data submitted by a client, inputting a mark generation model for mark generation, and acquiring mark data containing a plurality of candidate marks; Constructing a mark relation diagram based on the mark data, generating an editing interaction network corresponding to the mark relation diagram and returning to the client; Determining a mark editing parameter according to an editing instruction submitted by the client through the editing interaction network; And calling the mark generation model based on the mark editing parameters to carry out editing interaction processing corresponding to the editing instruction.
18. 18. A model-based logo processing apparatus comprising: And a memory configured to store computer-executable instructions that, when executed, cause the processor to: Acquiring multi-mode data generated by a mark input by a user and submitting the multi-mode data to a server; Receiving and displaying an editing interaction network which is returned by the server and is generated by constructing a sign relation graph based on sign data, wherein the sign data is obtained by inputting the multi-mode data into a sign generation model for sign generation; and acquiring an editing instruction submitted by the user and submitting the editing instruction to the server so as to determine a mark editing parameter according to the editing instruction and call the mark generation model to carry out editing interaction processing.
19. 19. A computer readable storage medium storing computer executable instructions which when executed implement the steps of the method of claim 1 or claim 12.

Description

Mark processing method and device based on model Technical Field The present document relates to the field of image processing technologies, and in particular, to a method and an apparatus for processing a marker based on a model. Background Along with the continuous development of artificial intelligence technology, the application of artificial intelligence in various fields is continuously expanded, intelligent auxiliary tools are widely applied to related processing systems, along with the increasing demands of users for diversification and individuation of image processing, related processing of images or visual elements is more and more concerned, such as sign processing or identification processing, under the condition, along with the continuous improvement of the requirements of users on various aspects of interaction convenience, intuitiveness and the like, how to realize better related processing for images or signs becomes a focus of attention of all parties. Disclosure of Invention One or more embodiments of the present disclosure provide a method for processing a flag based on a model, including obtaining multi-modal data submitted by a client and inputting a flag generation model to generate a flag, so as to obtain flag data including a plurality of candidate flags. And constructing a mark relation diagram based on the mark data, generating an editing interaction network corresponding to the mark relation diagram, and returning to the client. And determining mark editing parameters according to editing instructions submitted by the client through the editing interaction network. And calling the mark generation model based on the mark editing parameters to carry out editing interaction processing corresponding to the editing instruction. One or more embodiments of the present disclosure provide another method for processing a flag based on a model, which includes obtaining multi-mode data input by a user for generating a flag and submitting the multi-mode data to a server. And receiving and displaying the editing interaction network which is returned by the server and is generated by constructing a sign relation graph based on sign data, wherein the sign data is obtained by inputting the multi-mode data into a sign generation model to generate signs. And acquiring an editing instruction submitted by the user and submitting the editing instruction to the server so as to determine a mark editing parameter according to the editing instruction and call the mark generation model to carry out editing interaction processing. One or more embodiments of the present specification provide a model-based token processing apparatus, including a token generation module configured to obtain multimodal data submitted by a client and input a token generation model for token generation, to obtain token data including a plurality of candidate tokens. And the relation diagram construction module is configured to construct a mark relation diagram based on the mark data, generate an editing interaction network corresponding to the mark relation diagram and return to the client. And the parameter determining module is configured to determine a mark editing parameter according to an editing instruction submitted by the client through the editing interaction network. And the editing processing module is configured to call the mark generation model based on the mark editing parameters to carry out editing interaction processing corresponding to the editing instruction. One or more embodiments of the present specification provide another model-based token processing apparatus, including a data submitting module configured to obtain multi-modal data input by a user for token generation and submit the multi-modal data to a server. The interactive network display module is configured to receive and display the edited interactive network which is returned by the server and is generated by constructing a sign relation graph based on sign data, wherein the sign data is obtained by inputting the multi-mode data into a sign generation model for sign generation. The instruction submitting module is configured to acquire an editing instruction submitted by the user and submit the editing instruction to the server so as to determine a mark editing parameter according to the editing instruction and call the mark generation model to carry out editing interaction processing. One or more embodiments of the present specification provide a model-based token processing apparatus including a processor and a memory configured to store computer-executable instructions that, when executed, cause the processor to obtain multimodal data submitted by a client and input a token generation model for token generation to obtain token data comprising a plurality of candidate tokens. And constructing a mark relation diagram based on the mark data, generating an editing interaction network corresponding to the mark relation diagram, and returning to the