AI-driven disease detection in guava plants using teachable machine learning models

Guava (Psidium guajava), a widely cultivated tropical fruit, is highly susceptible to various diseases such as anthracnose, rust, wilt and algal leaf spot, which can lead to significant yield losses and economic setbacks. Traditional methods of disease identification rely heavily on manual inspection, which is time-consuming, subjective and dependent on expert availability, thereby making timely intervention difficult, particularly in rural or resource-limited areas. The integration of artificial intelligence (AI) and image-based machine learning (ML) offers a promising solution to automate and enhance the accuracy of disease detection in plants. This study explores the use of Googles’ teachable machine (GTM), a no-code AI platform for developing a user-friendly, image-based classification model to identify guava leaf diseases. A dataset comprising 600 images, categorised into five classes (healthy, anthracnose, rust, algal leaf spot and wilt), was used to train the model. The model achieved an overall classification accuracy of 96.2 %, with class-wise accuracies ranging from 94.2 % to 97.5 %. Confusion matrix analysis indicated strong predictive capability, with minor misclassifications occurring between visually similar disease symptoms. The results validate the feasibility of using teachable machine for real-world agricultural applications, particularly in field-level diagnosis where accessibility to advanced tools or expertise is limited. This AI-driven approach not only enhances the accuracy and speed of disease detection but also empowers farmers and agricultural workers with a practical tool for timely crop health management. The study underscores the potential of accessible AI platforms in promoting precision agriculture and sustainable fruit production.