A display screen point line defect multi-model fusion detection method based on grpc

CN116681995BActive Publication Date: 2026-07-21FREESENSE IMAGE TECH
View PDF 1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
FREESENSE IMAGE TECH
Filing Date
2023-06-13
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing deep learning models suffer from accuracy loss when called across programming languages, single models are ineffective in detecting different types of defects, have stringent hardware requirements, and are difficult to achieve fast and accurate classification of good and defective products.

Method used

A multi-model fusion scheme based on gRPC is adopted. The model is deployed and the pre- and post-processing pipeline is combined through the gRPC server. The client manages the images and performs threshold and location filtering. It supports cross-programming language calls and multi-model customization, and uses multi-threading parallelism to accelerate the transmission of results.

Benefits of technology

It enables fast, accurate, and low-cost classification of good and defective products, improves detection accuracy, reduces hardware requirements, and supports flexible calling and efficient transmission of multiple models.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116681995B_ABST
    Figure CN116681995B_ABST
Patent Text Reader

Abstract

The application provides a display screen point line defect multi-model fusion detection method based on grpc, including the following steps: a server side opens a deployment called channel pipeline, and according to a current project to be detected, different channels are called; a client side sends a picture to be verified to the server side according to a requirement, and the client side accepts data transmitted by the server side to obtain a detection result, and the application can realize a method for quickly, accurately and low-cost classifying good products and defective products, which is extremely practical.
Need to check novelty before this filing date? Find Prior Art