Automatic Identification of Conodonts Based on Deep Learning

Abstract

The study of conodonts can promote people's understanding of the major events of extinction and resuscitation in geological history. The identification of conodonts requires rich expertise. The identification time is long, the efficiency is low, and sometimes the accuracy is difficult to guarantee, which brings a lot of inconvenience to the in-depth study and application of conodonts. In this paper, we propose a deep learning model based on CNN for automatic identification of conodonts. In order to optimize the CNN model, we put forward a hierarchical classification method to solve the species relation problem among different categories. Also, we use softmax classification loss function to reduce the impact of class imbalance and class overlap. In order to prevent model over-fitting, we also use some data enhancement and anti-over-fitting methods. This work provides a reference for the realization of intelligent and automatic identification of conodonts.