MYCOLab Ai Logo - gold and army green

Classification of Mycena and Marasmius Species Using Deep Learning Models: An Ecological and Taxonomic Approach

Summary

Researchers developed an artificial intelligence system to automatically identify mushroom species from the Mycena and Marasmius groups by analyzing photographs. Using advanced computer vision and machine learning techniques, they achieved 98.9% accuracy in classification. This technology could help scientists, conservationists, and nature enthusiasts quickly identify mushroom species in the field, supporting biodiversity research and conservation efforts.

Background

Fungi play critical roles in ecosystems and biodiversity. The genera Mycena and Marasmius comprise thousands of species with distinct morphological and ecological characteristics. Traditional fungal classification methods are time-consuming and require expert knowledge, limiting scalability for large datasets.

Objective

To develop a novel deep learning framework for accurate classification of seven macrofungi species from Mycena and Marasmius genera by integrating CNNs with self-organizing maps and Kolmogorov-Arnold Networks. The study aims to improve taxonomic accuracy and demonstrate the potential of deep learning in fungal identification.

Results

MaxViT-S achieved the highest accuracy of 98.9% across all models. CNN-SOM improved accuracy from 52.9% to 86.3%, while CNN-KAN achieved 76.3% accuracy. MaxViT-SOM and ensemble methods also demonstrated strong performance. Chi-square testing confirmed statistical significance of results, with MaxViT-S and ResNetV2-50 showing superior performance.

Conclusion

The study successfully demonstrates the effectiveness of deep learning models in macrofungal species classification, with MaxViT-S achieving state-of-the-art results. This is the first application of SOM and KAN layers for fungal taxonomy. Future work will optimize KAN architecture and expand datasets to include more fungal classes.
  • Published in:Sensors (Basel),
  • Study Type:Comparative Study,
  • Source: 10.3390/s25061642
Scroll to Top