Few-shot Classification based on CBAM and prototype network
Shuo Xin, Han Liu · 2022
In recent years, with the continuous progress and development of deep learning, computer vision problems have been solved to a large extent, but the problem of few-shot has also appeared. Deep learning requires a lot of training data, so there is still a lack of learning from few samples. This paper focuses on the image classification problem in the few-shot problem as the research object. Firstly, based on the residual network ResNet18, and adding the convolution attention mechanism on this basis, the feature extraction network Res-CBAMnet is constructed. Secondly, the prototype network is used as the classifier to study the influence of different metric methods on the classification results. Experimental results show that the improved network achieves 99.72% and 67.42% accuracy on the few-shot image classification benchmark datasets Omniglot and mini-imagenet respectively.