Image-Based Classification of Healthy and Infected Cacao Pods Using Google Teachable Machine

Cacao cultivation is crucial for the global chocolate industry, yet it faces significant challenges from diseases such as Black Pod Rot and Pod Borer, which threaten yield and product quality. These issues are particularly pressing in the Philippines, where cacao production has grown steadily but remains vulnerable to these diseases. This study evaluates the effectiveness of Google Teachable Machine, a user-friendly webbased platform, in classifying healthy and diseased cacao pods. A dataset of 4,493 images [7]—comprising 3,344 healthy pods, 943 Black Pod Rot-infected pods, and 206 Pod Borer-infected pods—was used to train the machine learning model. Preprocessing steps included resizing all images to 224×224 pixels and normalizing pixel values to ensure consistent input quality. Transfer learning with MobileNet served as the backbone architecture, leveraging its lightweight and efficient design for image classification. The dataset was divided into training (85%) and testing (15%) sets to evaluate the model's performance on unseen data. Evaluation metrics included accuracy, precision, recall, and F1-score. The model achieved an overall accuracy of 91.11%, with precision at 96.85%, recall at 93.61%, and an F1score of 95.20%, underscoring its robust performance in distinguishing healthy and infected cacao pods. Slight overfitting was observed after 35 epochs, reflected in an increase in validation loss, highlighting the need for dataset expansion and regularization techniques. Despite these limitations, the model demonstrated significant potential for early disease detection. Future work should focus on increasing dataset diversity, particularly for underrepresented diseases like Pod Borer. Integrating the trained model into mobile applications would enable real-time disease diagnosis in field conditions, offering farmers a practical, cost-effective solution for improving disease management and promoting sustainable cacao production.