Increasing the Accuracy of a Neural Network Using Frequency Selective Mesh-to-Grid Resampling

TL;DR

This paper introduces a frequency selective mesh-to-grid resampling method to improve neural network accuracy by enhancing input image quality, demonstrating significant gains in classification performance across different architectures and datasets.

Contribution

The paper proposes a novel keypoint agnostic FSMR method for image interpolation that improves neural network training and inference accuracy compared to standard methods.

Findings

FSMR outperforms common interpolation in PSNR

FSMR increases classification accuracy by up to 4.31 percentage points

Application of FSMR benefits both training and inference phases

Abstract

Neural networks are widely used for almost any task of recognizing image content. Even though much effort has been put into investigating efficient network architectures, optimizers, and training strategies, the influence of image interpolation on the performance of neural networks is not well studied. Furthermore, research has shown that neural networks are often sensitive to minor changes in the input image leading to drastic drops of their performance. Therefore, we propose the use of keypoint agnostic frequency selective mesh-to-grid resampling (FSMR) for the processing of input data for neural networks in this paper. This model-based interpolation method already showed that it is capable of outperforming common interpolation methods in terms of PSNR. Using an extensive experimental evaluation we show that depending on the network architecture and classification task the application of FSMR during training aids the learning process. Furthermore, we show that the usage of FSMR in the application phase is beneficial. The classification accuracy can be increased by up to 4.31 percentage points for ResNet50 and the Oxflower17 dataset.