TeaLeafVision: An Explainable and Robust Deep Learning Framework for Tea Leaf Disease Classification
Summary: arXiv:2604.07182v1 Announce Type: cross
Abstract: As the world’s second most consumed beverage after water, tea is not just a cultural staple but a global economic force of profound scale and influence. More than a mere drink, it represents a quiet negotiation between nature, culture, and the human desire for a moment of reflection. So, the precise identification and detection of tea leaf disease is crucial. With this goal, we have evaluated several Convolutional Neural Networks (CNN) models, among them three shows noticeable performance including DenseNet201, MobileNetV2, InceptionV3 on the teaLeafBD dataset. The teaLeafBD dataset contains seven classes, six disease classes, and one healthy class, collected under various field conditions reflecting real-world challenges. Among the CNN models, DenseNet201 has achieved the highest test accuracy of 99%. In order to enhance the model reliability and interpretability, we have implemented Gradient weighted Class Activation Mapping (Grad CAM), occlusion sensitivity analysis, and adversarial training techniques to increase the noise resistance of the model. Finally, we have developed a prototype in order to leverage the model’s capabilities in real-life agriculture. This paper illustrates the deep learning model’s capabilities to classify the disease in real-life tea leaf disease detection and management.
Introduction
The cultivation of tea is vital for millions of people around the globe, significantly influencing local economies and cultures. However, the presence of diseases in tea leaves can lead to substantial economic losses. Therefore, a robust system for identifying and classifying these diseases is essential. In recent years, deep learning has emerged as a powerful tool in agricultural applications, particularly for disease detection.
Methodology
In our study, we explored various Convolutional Neural Network (CNN) architectures to identify the most effective model for tea leaf disease classification. The models evaluated include:
- DenseNet201
- MobileNetV2
- InceptionV3
These models were trained on the teaLeafBD dataset, which is specifically designed for tea leaf disease classification and includes:
- Six classes of disease
- One healthy class
- Diverse data collected under various field conditions
Results
Our experiments revealed that DenseNet201 achieved the highest test accuracy of 99%, demonstrating its effectiveness in classifying tea leaf diseases. Additionally, we implemented several techniques to enhance the model’s reliability:
- Gradient weighted Class Activation Mapping (Grad CAM) for interpretability
- Occlusion sensitivity analysis to assess model robustness
- Adversarial training to improve noise resistance
Conclusion
The results of our study highlight the potential of deep learning models in agricultural applications, particularly in the accurate classification of tea leaf diseases. The development of a prototype system based on our findings could significantly aid farmers and agricultural professionals in making informed decisions, ultimately leading to healthier crops and improved economic outcomes.
Future Work
We aim to further refine our models and extend the dataset to include more diverse conditions and disease types. Future research will also focus on integrating the model into mobile applications for real-time disease detection in the field.