Related Experiment Video
Updated: Jul 1, 2026
03:31
End-To-End Deep Neural Network for Salient Object Detection in Complex Environments
Published on: December 15, 2023
DSA-DET: a tea disease detection algorithm based on dynamic spatial pyramid and polarized linear attention
Haoxiang Dai1, Jiaxin Lv1, Chenlu Sun2
1Electrical Engineering College, Heilongjiang University, Harbin, China.
Frontiers in Plant Science
|June 3, 2026
Summary
This study introduces DSA-DET, an improved deep learning model for intelligent tea disease detection. The new method enhances accuracy and real-time performance, offering a practical solution for agricultural applications.
Area of Science:
- Agricultural Science
- Computer Vision
- Machine Learning
Background:
- Traditional tea disease detection methods are inefficient and subjective.
- Existing deep learning models struggle with accuracy and real-time performance in complex environments.
Purpose of the Study:
- To develop an intelligent tea disease detection method with improved accuracy and real-time capabilities.
- To address the limitations of current deep learning approaches in variable agricultural settings.
Main Methods:
- An improved Real-Time Detection Transformer (DSA-DET) was developed.
- Key innovations include a dynamic attention spatial pyramid backbone, an encoder with polarized linear attention and parallel spatial enhancement, and an efficient upsampling module.
Main Results:
- The DSA-DET model achieved 94.73% precision, 89.65% recall, and 93.68% mAP50.
- It demonstrated significant improvements over the baseline RT-DETR-R18 model.
- The model maintains a lightweight scale (15.4M parameters) and achieves a fast detection speed (71.5 FPS).
Conclusions:
- The proposed DSA-DET method successfully enhances tea disease detection accuracy.
- It offers a practical and reliable technical solution by balancing model complexity and inference speed for intelligent tea disease diagnosis.
