Sample image generation method, deep learning model training method and device

By segmenting the original image and determining the label of the region proportion of the original image, and generating sample images, the problem of insufficient model accuracy in multi-object image processing in the prior art is solved, and more efficient deep learning model training and multi-object recognition effects are achieved.

CN114882283BActive Publication Date: 2025-03-21SHANDONG LONGDU INTELLIGENT TECHNOLOGY CO LTD
0 Cites 0 Cited by

Patent Information

Application Number
CN202210541103.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-05-17
Publication Date
2025-03-21
Estimated Expiration
2042-05-17

AI Technical Summary

Technical Problem

The prior art is difficult to effectively process images containing multiple objects when generating sample images and training deep learning models, resulting in insufficient accuracy of the model in classification and object detection tasks.

Method used

By segmenting the original image, a label area is generated, and the label is determined based on the proportion of each object in the label area, and a sample image is generated. These sample images are used to train deep learning models to improve their classification and detection performance in multi-object scenarios.

Benefits of technology

By generating more accurate sample images, the performance of deep learning models in multi-object image classification and object detection tasks is improved, allowing the model to more accurately identify and classify images containing multiple objects.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present disclosure provides a sample image generation method, which relates to the field of artificial intelligence technology, and in particular to the field of deep learning, computer vision and autonomous driving technology. The specific implementation scheme is: performing image segmentation on an original image including at least one object to obtain at least one annotated area, wherein each object in the at least one object is surrounded by an annotated area; for at least one annotated area, determining at least one label according to the proportion of each object in the annotated area used to surround each object; and generating a sample image according to at least one label and the original image. The present disclosure also provides a deep learning model training method, object classification method, target object detection method, device, electronic device and storage medium.
Need to check novelty before this filing date? Find Prior Art