Search

• Empirical observations on convolutional neural networks are presented. • The size of the receptive field did not guarantee improved accuracy. • We observed that dead pixels exist in modern convolutional neural networks. • The effect of dead pixels is discussed with a solution to resolve them. Deep neural networks have been used in various fields, but their internal behavior in how they understand images is not well known. In this study, we discuss two counterintuitive properties of convolutional neural networks (CNNs). First, we evaluated the size of the receptive field of CNNs with their classification accuracy. Previous studies have attempted to increase the size of the receptive field for performance gain. However, we observed that some CNNs with a smaller receptive field can achieve higher classification accuracy. In this regard, we claim that a larger receptive field does not guarantee improved classification accuracy. Second, using the effective receptive field, we examined the contribution of each pixel to the output of CNN. Intuitively, each pixel is expected to equally contribute to the final output, but we found that there exist pixels in a partially dead state with little contribution to the output. We reveal that the reason for dead pixels lies in even stride operations with odd-sized kernels in CNN and propose a kernel padding method to remove the dead pixels. We demonstrated the vulnerability of CNNs with dead pixels when we detect a noise or small box that is on dead pixels. Our findings on dead pixels should be understood and considered in practical applications of CNN. [ABSTRACT FROM AUTHOR]

Recently, various normalization layers have been proposed to stabilize the training of deep neural networks. Among them, group normalization is a generalization of layer normalization and instance normalization by allowing a degree of freedom in the number of groups it uses. However, to determine the optimal number of groups, trial-and-error-based hyperparameter tuning is required, and such experiments are time-consuming. In this study, we discuss a reasonable method for setting the number of groups. First, we find that the number of groups influences the gradient behavior of the group normalization layer. Based on this observation, we derive the ideal number of groups, which calibrates the gradient scale to facilitate gradient descent optimization. This paper is the first to propose an optimal number of groups that is theoretically grounded, architecture-aware, and can provide a proper value in a layer-wise manner for all layers. The proposed method exhibited improved performance over existing methods in numerous neural network architectures, tasks, and datasets. • We propose a method to determine the number of groups of group normalization. • A theoretical analysis of group normalization with activation function is provided. • The proposed method is validated against various practical problems. [ABSTRACT FROM AUTHOR]