Extension of Convolutional Neural Network with General Image Processing Kernels

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.

Implementation of two-step gradual reset scheme for enhancing state uniformity of 2D hBN-based memristors for image processing

“…The hardware convolution was performed for an emblem input image with 210 × 210 size using four different kernels, as shown in figure 2(c). The functionalities of Sobel, Gaussian, and embossing kernels were edge detection, noise reduction, and 3D effect, respectively [40][41][42]. By converting the output currents to pixel values in post-processing, corresponding features were extracted through hardware convolution.…”

Voxel-based sensor placement for additive manufacturing applications

“…Kernel, in image processing, is a 2D matrix or a mask used for different operations like blurring, sharpening and edge detection [36]. Such a matrix is shown in Fig.…”