Classification of Healthy and Diseased Silkworms using Ensemble Learning and CNN

Sericulture is the technique of rearing silkworms to produce the cocoons that are essential for the manufacturing of silk. Silkworm is an insect which is frequently contracted by infections while being raised, which led to a significant annual loss in the number of cocoons produced. This paper implements a method that uses two machine learning ensemble algorithms: Random Forest (RF), Light Gradient Boosting Machine (LGBM) and state of art Convolutional Neural Network (CNN) to classify healthy and diseased silkworm. Gray-level Co-occurrence Matrix (GLCM) texture features are provided as input for training both RF and LGBM classifiers. The CNN is trained with its auto generated features. CNN, LGBM and RF resulted in recall of 85%, 75% and 59%, the precision of 85%, 75% and 61%, the F1-score of 57%, 75% and 85% and an average accuracy of 85%, 75% and 59% for classification of healthy and diseased silkworms.