Edible and Poisonous Mushroom Classification using Convolution Neural Network (CNN)

Indonesia is geographically located on the equator which makes it a tropical country, so it has biodiversity, one of which is mushrooms. The activity of clearing wild mushrooms is the daily life of the Indonesian people. However, this activity is at risk due to ignorance or inaccuracy in seeing the types of poisonous mushrooms that are identical to edible mushrooms. To overcome this problem, several researchers have used machine learning algorithms to help classify types of edible mushrooms and types of poisonous mushrooms. The research methodology uses the Convolutional Neural Network (CNN) method, then the data that has been collected requires preprocessing before being used in the CNN model. These steps can include resizing images to have uniform sizes, normalizing pixel intensities, as well as dividing the dataset into subsets for training, validation, and testing. This data set was obtained from Kaggle, which contains various images of mushrooms with labels indicating whether they are edible or poisonous. By using data sets from these various sources, CNN models can be better trained to recognize different types of fungi based on their visual characteristics. The combination of data from these sources ensures that the model has high accuracy in classifying edible and poisonous mushrooms. Based on the experimental results, the RestNet50 model shows better performance compared to other models with an accuracy value of 100%. Including the prediction results, the ResNet50 model gives good prediction results with the highest probability value of 95.64% and is consistently using both transfer learning and fine-tuning techniques, including the use of data balance and imbalance data. Whereas the MobileNetV2 model shows better results when using fine-tuning techniques.