MYCOLab Ai Logo - gold and army green

Detection and classification of Shiitake mushroom fruiting bodies based on Mamba YOLO

Summary

Researchers developed an artificial intelligence system called Mamba-YOLO that can automatically detect and grade shiitake mushrooms for harvest. The system looks at images of mushrooms and identifies which ones are ready to pick based on their size, maturity, and surface texture characteristics. With 98.89% accuracy and fast processing speed of 8.3 milliseconds, this technology could help automate mushroom harvesting and reduce labor costs for farmers. The compact model design also allows it to be installed on robotic harvesting machines.

Background

Shiitake mushroom production has become a major global industry worth approximately 68 billion USD in 2024, with China accounting for over 90% of global production. Current harvesting methods rely predominantly on manual picking, which is labor-intensive, inefficient, and prone to missing optimal harvest times. Mechanized harvesting requires precise detection and classification systems to improve efficiency and product quality.

Objective

To develop and validate a novel detection and classification method based on Mamba-YOLO for automated detection and quality grading of shiitake mushrooms according to picking standards and grade specifications. The method aims to address challenges of small targets, dense growth, and complex backgrounds in shiitake mushroom harvesting.

Results

The Mamba-YOLO model achieved precision of 98.89%, recall of 98.79%, mAP@0.5 of 97.86%, and mAP@0.5-0.95 of 89.97% on the test set. Individual category classification accuracies ranged from 96.2% to 98.8%. Detection speed was 8.3 ms with compact model parameters of 6.1 M. The model significantly outperformed other algorithms including Faster R-CNN, YOLOv5, YOLOv6, YOLOv7, and YOLOv8 in precision and recall metrics.

Conclusion

The Mamba-YOLO model provides an effective technical solution for automated shiitake mushroom detection, maturity determination, and quality grading based on cap texture characteristics. Its lightweight design, high detection accuracy, and real-time performance make it suitable for deployment on harvesting robots to improve automation and efficiency in shiitake mushroom production. Future work should focus on expanding dataset diversity, testing on edge devices, and improving performance under extreme lighting conditions.
  • Published in:Scientific Reports,
  • Study Type:Technical Development and Validation Study,
  • Source: 10.1038/s41598-025-00133-z, PMID: 40307284
Scroll to Top