Intelligent Wild Mushroom Recognition Using YOLOv11n: A YOLOv11n-Based System for Accurate and Real-Time Identification

Abstract

Accurate identification of wild mushrooms remains a persistent challenge due to the high morphological similarity between edible and toxic species. Traditional manual methods are labor-intensive, subjective, and error-prone, while existing computational approaches often prioritize algorithmic performance over practical, user-friendly solutions. This study presents an integrated intelligent recognition system designed to bridge this gap. A dataset of over 8,000 images representing 40 common wild mushroom species is constructed, and a detection model based on the YOLOv11n architecture is developed using transfer learning to improve feature extraction and real-time performance. Experimental results show that the model has achieved a mean average precision (mAP50) of 91.8 % ± 0.6 %, which is 4.6 percentage points higher than the YOLOv5-based mushroom recognition system (87.2 %). The system completes image recognition in under 1 second on a standard CPU without GPU acceleration, effectively balancing accuracy and speed. Furthermore, an intuitive visual interface is implemented to enable image upload [1], automatic detection, and result export, facilitating accessibility for both experts and the public. The system offers a practical, reliable tool for ecological research, biodiversity monitoring, and food safety assurance, demonstrating the successful translation of advanced deep learning methods into real-world applications.