Abstract:
Fungal species classification is an important task with broad implications for medicine, agriculture, and ecology. Even though they work well, traditional identification ...Show MoreMetadata
Abstract:
Fungal species classification is an important task with broad implications for medicine, agriculture, and ecology. Even though they work well, traditional identification techniques frequently take a lot of time, knowledge, and resources. In order to improve the effectiveness and precision of fungal classification, this work explores the use of deep learning techniques, more especially transfer learning with pre-trained Convolutional Neural Networks (CNNs). Using six well-known pre-trained models, namely ResNet-50, ResNet-101, VGG16, VGG19, InceptionV3, and InceptionV4,we modify them to categorize pictures from the DeFungi dataset, which is a varied compilation of images featuring fungal species. Our approach comprises a thorough comparative analysis based on metrics for F1 score, accuracy, and precision to assess how well these models perform in the identification of fungal species. The outcomes show that there is a lot of promise for using these trained models in mycology, with some models exhibiting better performance that might facilitate the creation of automated systems for fungal identification. This work not only advances the field of biological classification by illuminating the strengths and weaknesses of different CNN architectures for fungal classification, but it also emphasizes how crucial it is to use cutting-edge AI methods to tackle biological problems. ResNet-101 was the most accurate with accuracy result of 93 percentile of the six pre-trained models that were examined (ResNet-50, ResNet-101, VGG16, VGG19, InceptionV3, and InceptionV4).Other evaluation metrics including precision was 0.93, recall was 0.90, and F1 score was 0.91. Our results point to a promising direction for future research in automated fungal classification and provide insightful information for researchers and practitioners wishing to use machine learning in the identification and study of fungal species.
Date of Conference: 03-05 June 2024
Date Added to IEEE Xplore: 14 October 2024
ISBN Information: