Enhancing Papaya Disease Detection using Deep Learning and Data Augmentation Techniques: A Case Study in Sri Lanka

All farmers, particularly fruit growers, must take plant disease seriously. A significant problem for large farms is the spread of diseases that make the fruits unfit for consumption, thereby negatively impacting the farmer’s income. In order to stop the spread of disease, farmers must discover it early in its life cycle. Traditional fruit disease detection and identification rely on a person’s ability to see the diseased fruit. Even though this approach may suffice for small-scale farmers, it requires a high skill level for accurate identification. Machine Learning and Image Processing techniques have been used in recent research to develop computerised solutions to this challenge. This study considers Papaya, a popular fruit in Sri Lanka which suffers from a high post-harvest losses. Anthracnose, black spot, powdery mildew, phytophthora, and ringspot were selected among various papaya fruit diseases. They were choosen because they are the most prevalent papaya diseases in Sri Lanka. Data were collected from public image sources from the internet and actual fields. VGG 16, as a Convolutional Neural Network technique, was used to develop a computerised model for detecting papaya diseases. However, due to insufficient data, most image-based disease recognition systems have some limitations. However, novel data augmentation methods have promising advantages in expanding the limited data. This approach used Deep Convolutional Generative Adversarial Network (DCGAN) to expand the data set. The VGG 16 model accuracy was found using the same pre-processed data set with and without subjected to DCGAN. According to the results, the VGG 16 model has shown high accuracy across diseases. The resulting accuracy for Anthracnose, black spot, powdery mildew, phytophthora, and ringspot were 90%, 85%, 70%, 65% and 90%, respectively. The results indicate that the proposed DCGAN model performs better than basic techniques.