Deep Learning-Based Coconut Pest Detection and Size Classification System

Agriculture plays a vital role in sustaining economies, especially in countries like India where coconut
cultivation is significant. However, pest infestation and manual grading challenges reduce productivity and
efficiency. This paper proposes an integrated system combining Deep Learning (CNN + Autoencoder) for
pest detection and K-Nearest Neighbors (KNN) for coconut size classification. The system uses image
processing techniques and machine learning algorithms to automate pest identification and grading.
Experimental results show high accuracy (up to 94.5% for pest detection and 85% for size classification),
demonstrating the effectiveness of the proposed approach. Future enhancements include mobile-based
real-time detection and IoT integration.