Papaya (Carica papaya L.) is an important horticultural commodity whose production is often affected by post harvest diseases such as Anthracnose, Black Spot, and Powdery Mildew. Conventional visual inspection is generally subjective and may lead to inconsistent disease identification. To address this issue, this study proposes an automated papaya disease classification system based on Deep Learning. A modified Convolutional Neural Network (CNN) using the InceptionV3 architecture with Transfer Learning was implemented by incorporating Global Average Pooling and dual Dense layers with Dropout to improve feature extraction from papaya lesion images. The proposed model was evaluated using a dataset of 567 images and achieved a classification accuracy of 94.29%. To evaluate the model's generalization capability on a relatively limited dataset, 5-Fold Cross Validation was employed, resulting in an average accuracy of 92.61% with a standard deviation of ± 3.10%, indicating consistent performance across different data partitions. The proposed model also achieved higher accuracy than the comparative models, namely MobileNetV2 (91.43%) and ResNet50 (32.85%). Furthermore, the trained model was implemented in a web based application using Streamlit, achieving an average inference time of 0.1168 seconds per image. These results indicate that the proposed approach is capable of supporting web based papaya disease diagnosis with high classification performance and fast inference.
Copyrights © 2026