Advanced Felis Taxonomy Classification Using EfficientNetB3: A Deep Learning Approach

Accurate taxonomy of Felis species is necessary to understand their evolutionary linkages and thereby support ecological research and enhancement of conservation activities. Using a modified EfficientNetB3 model, this work classifies seven Felis species: African Wildcat, Blackfoot Cat, Chinese Mountain Cat, Domestic Cat, European Wildcat, Jungle Cat, and Sand Cat. The model was trained and evaluated using a 52-picture Kaggle dataset to produce an overall accuracy of 83%. Although the Blackfoot Cat and European Wildcat performed well, the model failed to accurately and with 100% recall classify the Sand Cat. Still, it wasn't easy to separate the African Wildcat, Chinese Mountain Cat, and Domestic Cat given their physical similarities. The study reveals that although the model performs well in many respects, more refining and better training data could help to increase classification accuracy for visually similar species. Combining deep learning with targeted data to support ecological research and conservation this approach boosts animal study.