PK-GCN: Prior Knowledge Assisted Image Classification using Graph Convolution Networks
This work addresses image classification challenges, especially in data-scarce scenarios, by integrating prior knowledge, but it is incremental as it builds on existing graph convolution and CNN methods.
The authors tackled the problem of image classification by incorporating class similarity knowledge into convolutional neural networks using a graph convolution layer, resulting in improved classification accuracy, particularly with small datasets, as demonstrated on MNIST and CIFAR10 benchmarks.
Deep learning has gained great success in various classification tasks. Typically, deep learning models learn underlying features directly from data, and no underlying relationship between classes are included. Similarity between classes can influence the performance of classification. In this article, we propose a method that incorporates class similarity knowledge into convolutional neural networks models using a graph convolution layer. We evaluate our method on two benchmark image datasets: MNIST and CIFAR10, and analyze the results on different data and model sizes. Experimental results show that our model can improve classification accuracy, especially when the amount of available data is small.