Fine-grained image recognition based on integrated transfer learning

Abstract

In order to improve the accuracy of fine-grained image recognition, a fine-grained image recognition method based on integrated transfer learning is proposed. This method uses two models of VGG-16 and inception-v3, combined with transfer learning to construct a K-selection transfer learning algorithm. The network model parameters pre-trained on the large-scale natural image dataset ImageNet are transferred to the transfer framework model of the fine-grained image dataset, and the network model is fine-tuned. The algorithm proposed in this paper is tested on CUB-200-2011 dataset, Oxford 102 Flower dataset, Stanford Cars dataset, FGVC-Aircraft dataset and Stanford Dogs dataset. The experimental results show that, compared with existing algorithms, the proposed algorithm has improved fine-grained image recognition accuracy under different models, and has optimized the adaptive effect of migration.