Automated Fungi Classification Using Deep Learning Based Approaches with Explainable AI

Fungal infections present a significant health problem worldwide, and the symptoms range from superficial to systemic. These infections characteristically cause high morbidity and mortality. Because fungus species have similar morphology, traditional diagnostic techniques like microscopic inspection and culture are often laborious and prone to errors. This work presents a deep learning-based method to improve fungal diagnoses' speed and precision. We used a number of cuttingedge deep learning models, such as DenseNet201, InceptionResNetV2, MobileNet, Xception, ResNetV2, VGG19, EfficientNetV2S and EfficientNetB0, using a dataset of 9,114 microscopic pictures from five main fungus families. After a thorough evaluation of these models' performance, Xception and EfficientNetV2S performed better than the others, with accuracy levels over 93.9% and F1 scores close to 95%. Our results show that deep learning has the potential to significantly shorten diagnosis times, which will speed up clinical judgment and therapy execution.