Extension of Convolutional Neural Network with General Image Processing Kernels

TL;DR

This paper introduces GFNN, a convolutional neural network variant that uses pre-defined image processing kernels in the first layer, reducing training time and achieving high accuracy with fewer samples.

Contribution

The paper proposes incorporating pre-defined, general-purpose image processing kernels into CNNs, demonstrating improved training efficiency and accuracy without specialized training.

Findings

Reduces training time by 30% compared to regular CNNs.

Achieves 90% accuracy with only 500 samples.

Reaches 99.56% accuracy on MNIST without ensembles or special algorithms.

Abstract

We applied pre-defined kernels also known as filters or masks developed for image processing to convolution neural network. Instead of letting neural networks find its own kernels, we used 41 different general-purpose kernels of blurring, edge detecting, sharpening, discrete cosine transformation, etc. for the first layer of the convolution neural networks. This architecture, thus named as general filter convolutional neural network (GFNN), can reduce training time by 30% with a better accuracy compared to the regular convolutional neural network (CNN). GFNN also can be trained to achieve 90% accuracy with only 500 samples. Furthermore, even though these kernels are not specialized for the MNIST dataset, we achieved 99.56% accuracy without ensemble nor any other special algorithms.