Optimizing Mushroom Classification using ML Classifier with MobileNet Features

Along with being a great source of nourishment, mushrooms also have important therapeutic qualities, including anti-inflammatory and immune-boosting actions. In this work, we present a robust analysis of mushroom classification using MobileNet features and six machine learning classifiers to classify five different mushroom classes (Amanita, Suillus, Hygrocybe, Entoloma, and Agaricus): KNN, Gaussian Naive Bayes (GNB), Random Forest (RF), Decision Tree (DT), LightGBM, and SVM. Results show that SVM and LightGBM achieved test-accuracy of 85.61% and 79.48% respectively and performed better than other models, highlighting the importance of feature extraction and model selection in reaching high classification accuracy. Automated mushroom identification has applications in various sectors to name a few: food safety, health, and ecology.