Fig. 1

Structure of the entire model consisting of a feature extractor for feature extraction, concatenation part for merging features extracted from each side image, and classifier for classification based on the combined features. PA posteroanterior, Lat lateral, FC fully connected, C channel size