Explainable Convolutional Neural Networks for the identification of the Ampullariidae genus

Invasive freshwater snails of the family Ampullariidae present significant threats to agriculture, biodiversity, and public health. Recent advances in deep learning have enabled automated identification of medically and agriculturally important snail species, but most existing models function as opaque "black boxes", limiting their utility in taxonomic research and ecological decision-making. In this study, we propose an explainable convolutional neural network (CNN) architecture based on VGG16 and Grad-CAM to classify and interpret morphological features of Ampullariidae and related genera. A curated dataset comprising 350 field-collected Ampullariidae specimens from northern Nigeria, augmented with labeled images of Biomphalaria, Bulinus, Lymnae, and Melanoides, was used to train and validate the model. The classifier achieved a validation accuracy of 0.99 within 20 epochs, indicating robust performance. Grad-CAM overlays revealed that the network correctly focused on genus-specific shell features, such as spire height, coiling direction, and aperture orientation. Our findings demonstrate that explainable deep learning can enhance taxonomic precision and provide visual insights into diagnostic traits, making it a powerful tool for ecological monitoring and parasite control programs.