PlantDiseaseNet-RT50: A Fine-tuned ResNet50 Architecture for High-Accuracy Plant Disease Detection Beyond Standard CNNs

arXiv ID: 2512.18500v1

Published: 2025-12-20

Authors: Santwana Sagnika, Manav Malhotra, Ishtaj Kaur Deol, Soumyajit Roy, Swarnav Kumar

Categories: cs.CV, cs.AI, cs.LG

Relevance Score: 0.70 / 1.00

View on arXiv Download PDF

Summary

PlantDiseaseNet-RT50 is a novel fine-tuned ResNet50 architecture designed for automated, high-accuracy plant disease detection, achieving approximately 98% accuracy, precision, and recall across various crop species. This computationally efficient deep learning solution addresses the significant threat of plant diseases to agricultural productivity, moving beyond traditional manual inspection methods to support global food security.

Medical Relevance

While focused on agriculture, this research is critically relevant to public health and global well-being by directly addressing food security and nutrition. Mitigating crop losses through accurate disease detection ensures stable food supplies, preventing malnutrition and food scarcity-related health crises that significantly impact human health.

AI Health Application

This AI application is an agricultural diagnostic tool designed for plant disease detection. While not a direct medical AI application for human or animal patients, its impact is directly linked to human health by ensuring global food security. By preventing significant crop losses, the AI indirectly supports human nutrition and well-being, thus serving as an upstream health intervention and a biosecurity measure against threats to the food supply chain.

Key Points

Introduces PlantDiseaseNet-RT50, a fine-tuned deep learning model based on ResNet50 for automated plant disease detection, addressing 70-80% crop losses worldwide.
The architecture incorporates strategically unfrozen terminal layers from the pretrained ResNet50 backbone, allowing for specialized feature extraction.
Features a custom classification head with integrated regularization mechanisms, including Batch Normalization and Dropout, to enhance generalization and prevent overfitting.
Employs dynamic learning rate scheduling via cosine decay, an advanced training technique designed to optimize model convergence and performance.
Achieves exceptional performance metrics, including approximately 98% accuracy, precision, and recall, on a comprehensive dataset covering distinct plant disease categories across multiple crop species.
Demonstrates that targeted fine-tuning of a standard pretrained CNN can transform it into a highly specialized and effective agricultural diagnostic tool.
Offers a computationally efficient solution for rapid and accurate plant disease diagnosis, designed for practical implementation in farming to enable timely interventions and reduce crop losses.

Methodology

The authors developed PlantDiseaseNet-RT50 by fine-tuning a pretrained ResNet50 deep learning model. Key modifications included strategically unfreezing terminal layers of the ResNet50 backbone and integrating a custom classification head. This custom head incorporated regularization techniques such as Batch Normalization and Dropout. The model was trained using dynamic learning rate scheduling facilitated by cosine decay on a comprehensive dataset encompassing various plant disease categories across multiple crop species.

Key Findings

PlantDiseaseNet-RT50 achieved exceptional diagnostic performance, demonstrating approximately 98% accuracy, precision, and recall in identifying distinct plant diseases. This high level of performance validates the effectiveness of the proposed architectural modifications, targeted fine-tuning strategy, and optimized training protocol in transforming a standard CNN into a specialized agricultural diagnostic tool.

Clinical Impact

The ability of PlantDiseaseNet-RT50 to provide rapid and accurate plant disease diagnoses has a profound indirect clinical and practical impact. By significantly reducing crop losses and enhancing agricultural productivity, it directly contributes to global food security. This, in turn, ensures more stable and affordable food supplies, which are fundamental for public health by preventing malnutrition, improving dietary quality, and safeguarding against health crises stemming from food scarcity, especially in vulnerable populations.

Limitations

The abstract does not explicitly state any limitations of the PlantDiseaseNet-RT50 model or the study design. It focuses primarily on the achievements and positive implications.

Future Directions

The abstract does not explicitly mention specific future research directions, though the implication is continued development and wider deployment of such AI-driven agricultural tools.

Medical Domains

Public Health (via food security) Nutritional Epidemiology Global Health Initiatives Preventive Medicine (indirectly, through stable food supply)

Keywords

Plant Disease Detection Deep Learning ResNet50 Fine-tuning Agricultural AI Computer Vision Food Security Global Health

Abstract

Plant diseases pose a significant threat to agricultural productivity and global food security, accounting for 70-80% of crop losses worldwide. Traditional detection methods rely heavily on expert visual inspection, which is time-consuming, labour-intensive, and often impractical for large-scale farming operations. In this paper, we present PlantDiseaseNet-RT50, a novel fine-tuned deep learning architecture based on ResNet50 for automated plant disease detection. Our model features strategically unfrozen layers, a custom classification head with regularization mechanisms, and dynamic learning rate scheduling through cosine decay. Using a comprehensive dataset of distinct plant disease categories across multiple crop species, PlantDiseaseNet-RT50 achieves exceptional performance with approximately 98% accuracy, precision, and recall. Our architectural modifications and optimization protocol demonstrate how targeted fine-tuning can transform a standard pretrained model into a specialized agricultural diagnostic tool. We provide a detailed account of our methodology, including the systematic unfreezing of terminal layers, implementation of batch normalization and dropout regularization and application of advanced training techniques. PlantDiseaseNet-RT50 represents a significant advancement in AI-driven agricultural tools, offering a computationally efficient solution for rapid and accurate plant disease diagnosis that can be readily implemented in practical farming contexts to support timely interventions and reduce crop losses.

Comments

This work is published in 2025 IEEE International Conference on Advances in Computing Research On Science Engineering and Technology (ACROSET). 6 pages, 2 figures, 2 tables

Journal Reference

2025 IEEE International Conference on Advances in Computing Research On Science Engineering and Technology (ACROSET), INDORE, India, 2025, pp. 1-6