Image recognition method and device and electronic equipment

By intercepting the region of interest from the target image and inputting the deep learning model, the feature similarity is calculated to determine the recognition result, the problem of automatic image recognition without the need for relevant training samples is solved, and the recognition effect with high accuracy and low cost is achieved.

CN119762884APending Publication Date: 2025-04-04BEIJING QIYI CENTURY SCI & TECH CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411958016.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-12-30
Publication Date
2025-04-04

AI Technical Summary

Technical Problem

The prior art is difficult to achieve effective training of deep learning models and automated image recognition without the need to utilize training samples related to the object to be identified, especially in the case where data has a long tail distribution, such as difficulty in collecting tattoo images.

Method used

The first modal feature is generated by intercepting the region of interest from the target image and inputting it into the target pretrained model belonging to the deep learning model. Then, the similarity between the first modal feature and the second modal feature corresponding to each prompt word in the prompt phrase is calculated, and the recognition result of the target sub-image is determined, thereby realizing automatic recognition of the presentation state of the target object.

Benefits of technology

Without the need to utilize training samples related to the target object, automated recognition of images containing the target object is achieved, improving the accuracy of recognition, and shortening business development time and cost.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119762884A_ABST
    Figure CN119762884A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides an image recognition method and device and electronic equipment, and relates to the technical field of image recognition. The method comprises the following steps: intercepting a target image to obtain a target sub-image; inputting the target sub-image into a target pre-training model belonging to a deep learning model, so that the target pre-training model generates a vector feature of a predetermined dimension of the target sub-image to obtain a first modal feature; calculating the similarity between the first modal feature and a second modal feature corresponding to each cue word in the cue word group to obtain the similarity corresponding to each cue word; based on the similarity corresponding to each cue word, determining an identification result corresponding to the target sub-image; and determining a presentation state of the target image to the target object based on the recognition result corresponding to the target sub-image. According to the scheme, the image can be automatically recognized under the condition that a training sample related to the object to be recognized does not need to be used for carrying out model training on the deep learning model.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Cited By

  • Target identification method and system

    CN120182586A

  • Target recognition method and system

    CN120182586B