A knowledge distillation method for remote sensing target detection based on feature separation attention

By employing a knowledge distillation method for remote sensing target detection based on feature separation attention, the foreground and background regions of remote sensing images are separated. An attention mask is calculated and L2 distillation loss is utilized, thereby improving the detection accuracy and performance of a lightweight remote sensing target detection network.

CN114565045BActive Publication Date: 2026-04-07BEIHANG UNIV
1 Cites 0 Cited by

Patent Information

Application Number
CN202210194931.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-03-01
Publication Date
2026-04-07
Estimated Expiration
2042-03-01

AI Technical Summary

Technical Problem

Existing lightweight deep learning object detection algorithms lack sufficient detection accuracy in remote sensing images, and existing knowledge distillation methods have unclear information definitions in the field of object detection, resulting in severe interference from background regions and failing to meet practical needs.

Method used

A knowledge distillation method for remote sensing target detection based on feature separation attention is adopted. By extracting feature maps of teacher and student networks, foreground and background regions are separated, and foreground and background attention masks are calculated. The knowledge of teacher network is transferred to student network using L2 distillation loss, thereby improving detection accuracy.

Benefits of technology

Without increasing computational costs, the detection accuracy of the lightweight object detection network is significantly improved, the false negative and false positive rates are reduced, and the model's discriminative ability is enhanced.

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

Abstract

This invention discloses a knowledge distillation method for remote sensing target detection based on feature-separated attention, comprising: extracting feature attention maps from the feature maps output by the teacher network and the student network respectively; separating the foreground and background regions of the feature maps, and calculating a foreground attention mask using the feature attention map of the teacher network and a background attention mask using the feature attention map of the student network; calculating the L2 distillation loss using the foreground and background attention masks; and transferring the knowledge from the teacher network to the student network based on the L2 distillation loss. This invention can effectively select the region to be distilled, improve distillation efficiency, and improve the detection accuracy of the final lightweight target detection network without changing the student network structure or increasing computational cost.
Need to check novelty before this filing date? Find Prior Art