Lightweight target detection method and device based on task interaction

By introducing a lightweight task interaction structure into the object detection model, the number of model parameters and calculation complexity are reduced, and the problems of large number of model parameters and high calculation amount in the prior art are solved, efficient object detection is achieved and hardware requirements and training costs are reduced.

CN119723151AActive Publication Date: 2025-03-28GUANGZHOU UNIVERSITY
4 Cites 0 Cited by

Patent Information

Application Number
CN202411587471.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-08
Publication Date
2025-03-28
Estimated Expiration
2044-11-08

AI Technical Summary

Technical Problem

The existing open source object detection model has large model parameters and high computational volume due to the output of the feature pyramid or the use of upper-level features of the classification branch/positioning branch, which increases hardware requirements and actual application costs.

Method used

A lightweight object detection method based on task interaction is proposed. Through the model composed of backbone network, neck network, task interaction lightweight structure and detection head, the number of model parameters is reduced and the detection accuracy is maintained. The specific steps include acquiring image data, performing feature extraction, fusion and task interaction processing, generating lightweight task feature data, and inputting a detection head for prediction.

Benefits of technology

While reducing the amount of model parameters, it maintains a high target detection accuracy, reduces the computational complexity and memory usage, improves the model's inference speed, reduces hardware requirements, and reduces the model's training cost.

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

Abstract

The invention discloses a lightweight target detection method and device based on task interaction. The method comprises the following steps: acquiring image data; inputting the image data into a backbone network for feature extraction processing to obtain image basic feature data; inputting the image basic feature data into a neck network for feature fusion processing to obtain image fusion feature data; inputting the image fusion feature data into a task interaction lightweight structure for task interaction processing to obtain lightweight task feature data; the lightweight task feature data comprises classification task features and positioning task features; and inputting the lightweight task feature data into a detection head for prediction to obtain a lightweight target detection result. The method can maintain high target detection precision while reducing the model parameter quantity, and can be widely applied to the technical field of data processing.
Need to check novelty before this filing date? Find Prior Art