Sample image generation method and device, electronic equipment and storage medium

CN119380137APending Publication Date: 2025-01-28CHINA AUTOMOTIVE INNOVATION CORP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411440288.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-15
Publication Date
2025-01-28

AI Technical Summary

Technical Problem

In the prior art, the method of generating sample images based on copy-paste or image layout is poor in quality and cannot effectively provide high-quality and diverse object detection data.

Method used

By obtaining the category information, coordinate information, and geometric conditions of the target object, the position grid network is constructed, and the coordinate information of the target object is converted into a grid -based label information, and the sample image is generated based on the diffusion model.

Benefits of technology

The quality and diversity of the sample image are improved, and all geometric controls are integrated into the diffusion model through text prompts to flexibly generate high -quality sample images to solve the problem of poor sample image quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119380137A_ABST
    Figure CN119380137A_ABST
Patent Text Reader

Abstract

The invention discloses a sample image generation method and device, electronic equipment and a storage medium, and relates to the technical field of image processing, and the generation method comprises the steps: obtaining target object information, sample image description information and geometric condition information of a to-be-generated sample image; determining a position grid of the sample image based on the sample image description information, and converting coordinate information of the target object into grid marking information based on the position grid; constructing geometric layout information for the target object based on the grid mark information and the category information of the target object; establishing a cue word based on the geometric layout information of the target object in the sample image and the geometric condition information of the sample image; and inputting the cue word into the diffusion model, and outputting a sample image. According to the invention, the technical problem that the quality of the generated sample image is poor in a method for generating the sample image based on copying and pasting or image layout in the prior art is solved.
Need to check novelty before this filing date? Find Prior Art