Fine-Grained Butterfly Recognition with Deep Residual Networks: A New Baseline and Benchmark

Thanks to the advances in deep learning techniques and the increasing size of training data, ground- breaking progress on image classification has recently been achieved. However, focusing on distinguishing usually hundreds of sub-categories belonging to the same basic-level category, fine- grained recognition of unusual natural object categories (e.g., a special type of insect) still remains challenging and needs to be solved. Due to mainly lack of sufficient annotated data, the state-of-the-art image classification approaches cannot well adapt to address the fine-grained challenges. In this paper, we study the problem of fine-grained butterfly recognition by introducing a new large-scale benchmark, which includes 82 butterfly categories. Moreover, we perform empirical study of the existing state-of-the-art image classification approaches and adopt ResNet as a new baseline. Extensive experiments under empirical settings demonstrate the superiority of the proposed baseline.