CDPNet: a deformable ProtoPNet for interpretable wheat leaf disease identification
Jinyu Zeng1, Bingjing Jia1, Chenguang Song1
1College of Information and Network Engineering, Anhui Science and Technology University, Bengbu, Anhui, China.
A new wheat leaf disease identification model, CDPNet, improves accuracy by leveraging contrastive learning and attention mechanisms. This advanced computer vision approach enhances disease detection in field conditions.
Area of Science:
- Agricultural Science
- Computer Vision
- Machine Learning
Background:
- Accurate wheat leaf disease identification is vital for food security.
- Existing computer vision models face challenges with scattered lesions and lack interpretability.
Purpose of the Study:
- To develop an interpretable computer vision model for wheat leaf disease identification.
- To improve disease recognition accuracy in field conditions.
Main Methods:
- Proposed the Contrastive Deformable Prototypical part Network (CDPNet).
- Utilized Cross Attention (CA) for enhanced feature discriminability.
- Employed Barlow Twins self-supervised contrastive learning to address data scarcity.
Main Results:
- Achieved an average recognition accuracy of 95.83% on the wheat leaf disease dataset.
- Outperformed the baseline model by 2.35%.
Conclusions:
- CDPNet offers superior performance for real-world wheat disease identification.
- The model provides clinically interpretable decision support.
More Related Videos
06:28Author Spotlight: Unraveling Plant Responses to Abiotic Stresses Using the PlantScreen Robotic Platform
Published on: June 7, 2024
07:36Visualizing Early Infection Sites of Rice Blast Disease Magnaporthe oryzae on Barley Hordeum vulgare Using a Basic Microscope and a Smartphone
Published on: March 17, 2023
