Investigation of an Efficient Multi-Class Cotton Leaf Disease Detection Algorithm That Leverages YOLOv11

Fangyu Hu, Mairheba Abula, Di Wang, Li Xuan, Ning Yan, Qing Xie, Xuedong ZhangSensors 2025 · DOI

TL;DR

This paper introduces ACURS-YOLO, a novel deep learning model for detecting multiple types of cotton leaf diseases. By integrating medical image segmentation techniques (U-Net v2) and attention mechanisms (CBAM) into the YOLOv11 architecture, ACURS-YOLO significantly improves accuracy and robustness in complex agricultural environments, outperforming existing models.

What Problem Does This Paper Address?

Cotton is a vital global crop, but leaf diseases can cause substantial yield losses. Traditional methods for identifying these diseases are labor-intensive, prone to errors due to visual similarity between diseases, and lack scalability. While deep learning, particularly Convolutional Neural Networks (CNNs), has shown promise, existing models still struggle with challenges like complex backgrounds, detecting small lesions in early stages, and generalizing well across different disease presentations and environmental conditions. This research aims to bridge this gap by developing a more accurate, efficient, and robust automated system for multi-class cotton leaf disease detection.

Methods

The core of this work is the ACURS-YOLO model, built upon the YOLOv11 architecture. Several key modifications and integrations were made:

  • U-Net v2 Integration: Inspired by its success in medical image segmentation, the U-Net v2 module was incorporated into the YOLOv11 backbone. This enhances multi-scale feature extraction, crucial for capturing the diverse visual characteristics of diseases.
  • CBAM Attention Mechanism: The Convolutional Block Attention Module (CBAM) was added to help the model focus on critical disease-related features and reduce interference from leaf texture and environmental noise.
  • SimSPPF Module: The standard Spatial Pyramid Pooling-Fast (SPPF) module was replaced with SimSPPF to reduce computational complexity and improve inference speed while maintaining multi-scale feature extraction.
  • C3k2_RCM Module: This module was appended to the neck network to enhance long-range contextual modeling and improve the detection of multi-scale targets in complex scenarios.
  • ARelu Activation Function: The ARelu activation function was introduced to mitigate the vanishing gradient problem, leading to more stable training and improved detection integrity.

Datasets

The authors constructed a specialized dataset for this study.

  • ACURS-YOLO Dataset: This dataset comprises 3000 images of six common cotton leaf diseases: Cotton Leaf Spot, Cotton Leaf Curl, Cotton Brown Spot, Cotton White Mold, Cotton Verticillium Wilt, and Cotton Fusarium Wilt. The images were sourced from online datasets and field photographs, with data augmentation techniques applied to enhance robustness. The dataset was split into training (80%), validation (10%), and testing (10%) sets.

Key Insights and Ideas

  • Medical Segmentation for Agriculture: The successful application of U-Net v2, a model renowned for medical image segmentation, highlights the potential for cross-domain transfer of advanced computer vision techniques to agricultural challenges.
  • Attention is Key: The CBAM module's ability to dynamically focus on disease regions significantly boosts detection accuracy, demonstrating the power of attention mechanisms in filtering out irrelevant information.
  • Efficiency Through Simplification: Replacing SPPF with SimSPPF and optimizing activation functions (ARelu) shows that computational efficiency and speed can be improved without sacrificing accuracy, which is critical for real-time applications.
  • Synergistic Module Effects: Ablation studies revealed that while each individual module contributes positively, their combined integration in ACURS-YOLO leads to a synergistic effect, yielding the best overall performance.

Key Results

The ACURS-YOLO model achieved impressive performance metrics:

  • Accuracy:
    • mAP_0.5: 94.6%
    • mAP_0.5:0.95: 83.4%
    • Precision: 95.5%
    • Recall: 89.3%
    • F1 Score: 92.3%
  • Speed: 148 frames per second (FPS), which is only slightly lower than the baseline YOLOv11 (156 FPS) but significantly better than many other models.
  • Comparison: ACURS-YOLO outperformed YOLOv11 and other conventional models (SSD, YOLOv8, YOLOv9, YOLOv10) across key evaluation metrics.
  • Complex Scenarios: In test scenarios with complex backgrounds, ACURS-YOLO showed a significant improvement in accurately recognizing diseases, with an average accuracy of 94.2% compared to YOLOv11's 75.0%.

Limitations and Future Work

  • Limitations:
    • Despite improvements, misclassification can still occur in extreme cases of occlusion, highly similar disease presentations, or very poor lighting conditions.
  • The dataset, while comprehensive, only covers six common diseases and doesn't represent disease variations across different cotton varieties (e.g., insect-resistant or long-staple cotton), which could affect generalizability.
  • Future Work:
    • Further optimization of the model architecture by exploring more advanced feature extraction and fusion methods, potentially integrating other attention mechanisms or novel neural network structures.
  • Expanding the dataset's scale and diversity to include more disease types, growing conditions, and cotton varieties to enhance generalizability.
  • Deploying the model on mobile terminals and integrating additional sensor data (spectral, multimodal) to create more robust and reliable disease detection systems for agricultural applications.

Summary updated

Abstract

Cotton leaf diseases can lead to substantial yield losses and economic burdens. Traditional detection methods are challenged by low accuracy and high labor costs. This research presents the ACURS-YOLO network, an advanced cotton leaf disease detection architecture developed on the foundation of YOLOv11. By integrating a medical image segmentation model, it effectively tackles challenges including complex background interference, the missed detection of small targets, and restricted generalization ability. Specifically, the U-Net v2 module is embedded in the backbone network to boost the multi-scale feature extraction performance in YOLOv11. Meanwhile, the CBAM attention mechanism is integrated to emphasize critical disease-related features. To lower the computational complexity, the SPPF module is substituted with SimSPPF. The C3k2_RCM module is appended for long-range context modeling, and the ARelu activation function is employed to alleviate the vanishing gradient problem. A database comprising 3000 images covering six types of cotton leaf diseases was constructed, and data augmentation techniques were applied. The experimental results show that ACURS-YOLO attains impressive performance indicators, encompassing a mAP_0.5 value of 94.6%, a mAP_0.5:0.95 value of 83.4%, 95.5% accuracy, 89.3% recall, an F1 score of 92.3%, and a frame rate of 148 frames per second. It outperforms YOLOv11 and other conventional models with regard to both detection precision and overall functionality. Ablation tests additionally validate the efficacy of each component, affirming the framework's advantage in addressing complex detection environments. This framework provides an efficient solution for the automated monitoring of cotton leaf diseases, advancing the development of smart sensors through improved detection accuracy and practical applicability.
BibTeX
@inproceedings{hu2025investigation,
  author = {Fangyu Hu and Mairheba Abula and Di Wang and Li Xuan and Ning Yan and Qing Xie and Xuedong Zhang},
  title = {{Investigation of an Efficient Multi-Class Cotton Leaf Disease Detection Algorithm That Leverages YOLOv11}},
  booktitle = {Sensors},
  year = {2025},
  doi = {10.3390/s25144432},
}
Figure