Related Experiment Video
Updated: Aug 9, 2026
07:14
Dual Raster-Scanning Photoacoustic Small-Animal Imager for Vascular Visualization
Published on: July 15, 2020
4.2K
WeedSwin hierarchical vision transformer with SAM-2 for multi-stage weed detection and classification
Taminul Islam1,2, Toqi Tahamid Sarker3,4, Khaled R Ahmed3,4
1School of Computing, Southern Illinois University, Carbondale, IL, 62901, USA. taminul.islam@siu.edu.
Scientific Reports
|July 2, 2025
Summary
This study introduces advanced computer vision for automated weed detection and classification in precision agriculture. A novel WeedSwin Transformer architecture achieves superior performance across multiple growth stages, enabling sustainable farming.
Area of Science:
- Agricultural Science
- Computer Vision
- Deep Learning
Background:
- Precision agriculture demands automated weed identification for sustainable practices.
- Existing methods face challenges in detecting diverse weed species across various growth stages.
Purpose of the Study:
- To develop and evaluate a robust system for weed detection and classification across multiple growth stages.
- To introduce novel datasets and a specialized deep learning architecture for enhanced weed identification.
Main Methods:
- Creation of two large datasets (AWD, BWD) with 16 weed species across 11 growth stages.
- Preprocessing using traditional computer vision and the SAM-2 model for precise annotations.
- Evaluation of state-of-the-art object detection models and proposal of the WeedSwin Transformer architecture.
Main Results:
- The proposed WeedSwin Transformer achieved 0.993 mAP and 0.985 mAR at 218.27 FPS.
- WeedSwin outperformed existing architectures in detecting diverse weed species and growth stages.
- The approach demonstrated robustness in identifying challenging 'driver weeds' impacting crop productivity.
Conclusions:
- The WeedSwin architecture and extensive datasets advance automated weed identification in agriculture.
- This research supports efficient, sustainable weed management and reduced herbicide use.
- The findings pave the way for improved precision farming techniques and crop management.
