Using Deep Learning to Improve Detection of Citrus Leaf and Fruit Diseases

Greg Howard
15th April, 2025

Using Deep Learning to Improve Detection of Citrus Leaf and Fruit Diseases

To improve the deep learning models' diagnostic accuracy, the training dataset was expanded by applying augmentations—including rotation (b), flipping (c), brightness adjustment (d), scaling (e), and noise addition (f)—to original images, as shown with a citrus fruit affected by Black spot disease (a).

Image adapted from: Goyal et al. / CC BY (Source)

Key Findings

  • Researchers at JECRC University in Jaipur used advanced AI to detect citrus plant diseases from images
  • Their deep learning models InceptionV3 and DenseNet121 accurately identified diseases with over 99% success
  • These AI tools can help farmers quickly spot and manage diseases, reducing crop losses and improving fruit quality
Citrus fruits are a staple in diets worldwide, valued for their rich vitamin C content and other essential nutrients. However, citrus production faces significant challenges due to various diseases that affect the fruit's quality and yield, leading to substantial financial losses for farmers. Early detection of these diseases is crucial to prevent their spread and mitigate economic impacts. Recent advancements in image processing and computer vision have shown promise in developing tools for the timely identification of plant diseases[2]. A recent study conducted by researchers at JECRC University, Jaipur, Rajasthan, India[1] explored the effectiveness of deep learning models in classifying citrus diseases from images. The study focused on four prominent deep learning architectures: EfficientNetB0, ResNet50, DenseNet121, and InceptionV3. These models were tested on a dataset comprising 759 images categorized into nine classes, which included common diseases such as Black Spot, Canker, Greening, Scab, Melanose, as well as healthy examples of fruits and leaves. The researchers conducted extensive experiments to evaluate the performance of each model. InceptionV3 and DenseNet121 demonstrated exceptional accuracy, both achieving a test accuracy of 99.12%. These models also recorded high macro and weighted average F1-scores of approximately 0.986 and 0.991, respectively. These scores indicate that the models were highly precise and reliable in correctly identifying and classifying the various disease types across the majority of the classes. In contrast, ResNet50 and EfficientNetB0 showed moderate performance with test accuracies of 84.58% and 80.18%, respectively. The high accuracy achieved by InceptionV3 and DenseNet121 underscores the potential of advanced convolutional neural networks (CNNs) in the field of agricultural disease management. These models can analyze images of citrus plants to accurately detect and classify diseases, providing farmers and agricultural professionals with valuable tools for proactive disease management. This capability can lead to reduced crop losses and improved yield quality, addressing the critical issues faced by citrus growers. This study builds upon previous research that highlighted the importance of image datasets and machine learning algorithms in plant disease detection[2]. The dataset used in this research was similar to the one presented in earlier studies, which included images of citrus fruits and leaves from various regions, offering a comprehensive resource for training and testing machine learning models. Additionally, a survey of existing digital image processing techniques for plant disease detection emphasized the effectiveness of visible spectrum imaging in identifying disease symptoms on leaves and stems[3]. The current study extends this knowledge by applying state-of-the-art deep learning models to enhance the accuracy and reliability of disease classification. By leveraging deep learning, the researchers were able to automate the process of disease detection, which traditionally relied on manual inspection by experts. This automation not only speeds up the identification process but also reduces the dependence on specialized knowledge, making disease management more accessible to a broader range of farmers. The high performance of models like InceptionV3 and DenseNet121 suggests that these technologies can be integrated into practical applications, such as mobile apps or automated monitoring systems, to provide real-time disease diagnostics in orchard settings. Furthermore, the study's findings contribute to the ongoing efforts to digitize agriculture and implement precision farming techniques. Precision farming involves using technology to monitor and manage crops with high precision, optimizing inputs like water, fertilizers, and pesticides to improve crop health and productivity. Accurate disease detection is a critical component of precision farming, and the successful application of deep learning models in this study highlights a significant step forward in achieving this goal. The collaboration between JECRC University and institutions that provide comprehensive image datasets, similar to the Department of Computer Science at the University of Gujrat and the Citrus Research Center in Pakistan[2], plays a vital role in advancing this research. Access to high-quality, diverse datasets is essential for training robust machine learning models that can generalize well to different environments and disease manifestations. In conclusion, the study conducted by JECRC University demonstrates the effective use of deep learning models in classifying citrus diseases with high accuracy. By building on previous research and utilizing comprehensive image datasets, the study provides a viable solution for early disease detection in citrus plants. This advancement holds significant promise for improving agricultural practices, reducing economic losses, and ensuring the quality and availability of citrus fruits for consumers worldwide.

AgricultureBiotechPlant Science

References

Main Study

1) Integrating advanced deep learning techniques for enhanced detection and classification of citrus leaf and fruit diseases

Published 12th April, 2025

https://doi.org/10.1038/s41598-025-97159-0

Related Studies

2) A citrus fruits and leaves dataset for detection and classification of citrus diseases through machine learning.

https://doi.org/10.1016/j.dib.2019.104340

3) Digital image processing techniques for detecting, quantifying and classifying plant diseases.

https://doi.org/10.1186/2193-1801-2-660


Related Articles

An unhandled error has occurred. Reload 🗙