Deteksi Jamur Beracun dan Tidak Beracun Menggunakan CNN dan YOLO

Abstract

In Indonesia, which has a tropical climate there is a lot of wood which is very suitable for mushroom growth. However, some of these mushrooms are poisonous and should be avoided. By examining the morphological characteristics of the mushroom, such as the shape of the mushroom cap, color, smell, and other characteristics it is possible to identify between poisonous and non-poisonous mushrooms. However, some poisonous mushrooms have the same morphological characteristics as non-poisonous mushrooms, making it difficult for us to differentiate them when seen with the naked eye. As a solution to this problem, machine learning is needed to identify which mushrooms are poisonous and which are not. As in this research, Convolutional Neural Network (CNN) and You Only Look Once (YOLO) were used as methods for identification. The stages in this research include collecting datasets, modeling and training data, and the final process, namely deployment. CNN Method got an accuracy value of 55% and succeeded in identifying 5 correct and 5 incorrect images from the 10 images given. Meanwhile, using YOLOv5, we got a high accuracy value of 91% and succeeded in identifying 9 correct and 1 incorrect image out of the 10 images provided. From a comparison of the 2 methods, it was found that detecting poisonous and non-toxic mushrooms was better using the YOLO method than the CNN method