Related Experiment Video
Updated: Jul 13, 2026
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
Deep learning for apple leaf disease diagnosis: a comparative study with convolutional neural networks and
Soroush Toutounchian1, Amirhosein Zobeiri1, Alireza Rezaee2
1Department of Mechatronics, School of Intelligent Systems, College of Interdisciplinary Science and Technology, University of Tehran, Tehran, Iran.
Scientific Reports
|May 27, 2026
Summary
This study enhanced apple leaf disease diagnosis using advanced deep learning models like Vision Transformers and Convolutional Neural Networks. Optimized models achieved high accuracy, improving food security through early detection.
Area of Science:
- Agricultural Science
- Computer Science
- Plant Pathology
Background:
- Plant diseases significantly threaten global food security and agricultural yields.
- Early and accurate diagnosis of plant diseases is crucial for preventing food loss and ensuring economic stability.
Purpose of the Study:
- To investigate the efficacy of Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) for diagnosing apple leaf diseases.
- To enhance diagnostic performance through DropBlock regularization and optimized classifier thresholds.
Main Methods:
- Modified feature extraction layers of CNNs and ViTs with DropBlock, utilizing ImageNet pre-trained weights.
- Fine-tuned models on the Plant Pathology 2021 dataset for multi-label classification of five diseases and a healthy class.
- Optimized Dropout and DropBlock probabilities, and employed Swarm Optimization Algorithms for classifier thresholds.
Main Results:
- Optimal DropBlock probability was 0.05 and Dropout probability was 0.2.
- SwinV2T achieved 90.7% accuracy, while SwinV2S reached a 91.7% F1-score, outperforming other architectures.
- DropBlock regularization and optimized thresholds significantly improved model performance.
Conclusions:
- Recent deep learning architectures and optimization algorithms show superior performance in apple leaf disease diagnosis compared to older methods.
- The developed models demonstrate significant potential for accurate and efficient identification of apple leaf diseases, aiding in crop management and food security.