Poisonous Mushroom Detection using YOLOV5

Abstract

Mushroom is a nutritious food that is grown and consumed in many countries around the world. It is preferred for consumption due to its easy acquisition, the benefit to human health and taste. Although edible ones are beneficial for health, there are also poisonous types. It is difficult for people who are not familiar with this subject to distinguish which mushrooms are edible. Therefore, it will be useful to provide this process automatically. The study aims to identify poisonous mushrooms. In this context, a dataset containing the eight most poisonous mushroom species is created. The dataset created is trained with the fine tuning method using the pre-trained YOLOV5 algorithm. Precision, recall, and mAP metrics are used to demonstrate the performance of the method. The fine-tuned model enables the recognition of eight different types as 0.77 mean Average Precision.