In this study using transfer learning method, a model has been developed to classify pictures of Iran snakes in one of the 10 classes considered. Prediction accuracy of snake type and poisoning severity (in three classes named: venomous, semi-venomous, and non-venomous) using a limited dataset (composed of 174 snake pictures divided into 112 pictures of the train dataset, and 62 pictures of the test dataset) was the primary aim followed in this study. The pre-trained model that our transfer learning model was based on it was the EfficientNet model. After training the transfer learning model, it was evaluated using the sensitivity, specificity, precision, F1 score, and accuracy criteria. According to the results of this study, training phase of the model lasted about 60 minutes and its accuracy was 76%. Due to the fact that the model was trained on a regular laptop computer without a GPU, its performance was acceptable.
Ahmadi, I. (2025). Implementation of transfer learning to classify 10 species of Iran snakes. Biodiversity and Animal Taxonomy, 2(2), 306-319. doi: 10.22126/jbat.2025.10545.1042
MLA
Iman Ahmadi. "Implementation of transfer learning to classify 10 species of Iran snakes". Biodiversity and Animal Taxonomy, 2, 2, 2025, 306-319. doi: 10.22126/jbat.2025.10545.1042
HARVARD
Ahmadi, I. (2025). 'Implementation of transfer learning to classify 10 species of Iran snakes', Biodiversity and Animal Taxonomy, 2(2), pp. 306-319. doi: 10.22126/jbat.2025.10545.1042
VANCOUVER
Ahmadi, I. Implementation of transfer learning to classify 10 species of Iran snakes. Biodiversity and Animal Taxonomy, 2025; 2(2): 306-319. doi: 10.22126/jbat.2025.10545.1042
)