A fine-grained classification method based on self-attention Siamese network
He Can, Guowu Yuan, Hao Wu · 2021
Compared with other fine-grained image classifications, the classification of wild snakes is more difficult and complicated. This is because snakes have different postures, move very fast, and are often coiled. Judging and classifying according to the local characteristics of snakes is difficult. To solve this problem, this paper applies the self-attention mechanism to fine-grained wild snake image classification, to solve the problem of convolutional neural networks that focus on the local part and ignore the global information due to the deepening of the number of layers. Use Swin Transformer for transfer learning to obtain a fine-grained feature extraction model. To further study the performance of the self-attention mechanism in the field of meta-learning, this paper improves the feature extraction model to build a Siamese network and construct a classifier to learn and classify a small number of samples. Compared with other methods, this method reduces the time and space consumption caused by feature extraction, improves the accuracy and efficiency of meta-learning classification, and increases the autonomous learning of meta-learning.