Have a personal or library account? Click to login
An Improved RT-DETR-Based Object Detection Algorithm for UAV Aerial Photography Cover

An Improved RT-DETR-Based Object Detection Algorithm for UAV Aerial Photography

Open Access
|Dec 2025

Figures & Tables

Figure 1.

RT-DETR structure diagram
RT-DETR structure diagram

Figure 2.

Improved RT-DETR algorithm
Improved RT-DETR algorithm

Figure 3.

SCSA overall structure diagram. (a) Shows the structure diagram of SCSA, and (b) shows the structure diagram of MS-DWConv1d
SCSA overall structure diagram. (a) Shows the structure diagram of SCSA, and (b) shows the structure diagram of MS-DWConv1d

Figure 4.

CA-SHSA Module Structure
CA-SHSA Module Structure

Figure 5.

CARAFE Module Structure Diagram
CARAFE Module Structure Diagram

Figure 6.

Structural diagram of RepC3
Structural diagram of RepC3

Figure 7.

BFAM Structure Diagram
BFAM Structure Diagram

Figure 8.

MFRC3 Module Structure Diagram
MFRC3 Module Structure Diagram

Figure 9.

Target category and size distribution chart
Target category and size distribution chart

Figure 10.

mAP curve comparison chart
mAP curve comparison chart

Figure 11.

Visualization of experimental results comparison
Visualization of experimental results comparison

EXPERIMENTAL HYPERPARAMETERS

HyperparametersConfigurations
batch_size4
epoch200
optimizerAdamW
num_workers4
image_size640×640

The effects of each improvement module

BasicSCSABlockCARAFEMFRC3Param/MBAPs/%APm/%APl/%mAP@0.5/%mAP@0.5:0.95/%FPS
---19.219.239.354.848.028.581.4
--18.119.339.454.848.128.680.5
--18.219.339.557.148.528.970.8
--34.220.342.056.650.530.449.5
-18.218.938.953.247.628.374.7
-34.320.541.753.351.030.145.6
-27.921.042.352.850.029.563.2
28.021.042.458.351.131.053.7

Comparison with mainstream algorithms

modelParam/MBGFLOPsFPSmAP@0.5%mAP@0.5 : 0.95%
YOLO v5x86.2203.83442.525.2
YOLO v8n3.08.111933.517.8
YOLO v8s11.228.78537.321.6
YOLO v10n2.698.211129.917.1
YOLO v10s8.0424.514336.421.4
Defor mable401962942.227.1
DETR DINO472792446.229.4
RT-DETR19.230.281.448.028.5
Ours28.056.053.751.131.0
Language: English
Page range: 59 - 69
Published on: Dec 31, 2025
In partnership with: Paradigm Publishing Services
Publication frequency: 4 issues per year

© 2025 Yingying Long, Liuhua Di, Yanfang Fu, Shifeng Zhao, Xiaojun Bai, published by Xi’an Technological University
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.