Performance Enhancement of Mushroom Species Classification via Modified InceptionV3

Abstrak

Mushrooms encompass a very large number of species, and some of them are toxic to humans. It is very difficult to classify mushroom species quickly and accurately, especially for common individuals who often encounter wild mushrooms in nature. To address this problem, this study envisioned an automated mushroom species classification system using deep learning methods and the InceptionV3 model. This model was chosen because it is highly generalizable, performs well with challenging images, and is precise for most image-based classification tasks. The dataset comprises 18 mushroom species and was created from a Kaggle version. Data balancing, preprocessing, data augmentation, and model training constitute the research work. The dataset has been divided into 70% training, 15% validation, and 15% test. The training results show that the model achieves 81.35% accuracy in identifying mushroom species. The study contributes to the development of AI-based image recognition technology that can help humans find mushrooms more rapidly and securely.