Union-net: lightweight deep neural network model suitable for small data sets

Traditional deep learning models prefer large data sets, and in reality small sets are easier to obtain. It is more practical build suitable for sets. Based on CNN, this paper proposes the concept of union convolution a model Union-net that The has size superior performance. In paper, tested based multiple commonly used experimental results show outperforms most when dealing with datasets, other complex classification tasks or few-shot datasets. codes have been uploaded https://github.com/yeaso/union-net .