Butterfly Species Classification with EfficientNetB1 Transfer Learning

Butterflies, as vital pollinators and ecological indicators, play a significant role in maintaining biodiversity and ecosystem health. However, their survival is increasingly threatened by habitat loss and climate change, necessitating accurate species classification for effective conservation. This study aims to propose a classification method based on the principle of transfer learning, which employs the EfficientNetB1 model to achieve effective classification of multiple butterfly species. Leveraging a publicly available dataset comprising 6,499 images across 75 butterfly species, the model was fine-tuned to optimize performance. The dataset was divided into training, validation, and test sets with a ratio of 8: 1: 1, with stratified sampling ensuring balanced class distribution. The research findings reveal that the EfficientNetB1 model attains a classification accuracy of 91.4% on the test set. This study highlights the significant potential of deep learning techniques, particularly the EfficientNetB1 model, in advancing the identification and conservation of butterfly species. Furthermore, it emphasizes the critical role of such approaches in enhancing the understanding and application of deep learning for biodiversity monitoring and ecological preservation.