The impact of resolution and flipping on image classification

Abstract

The initial stage of image classification involves image preprocessing, which includes considerations such as the object's position within the image, the background, ambient lighting, and the viewing angle. These factors can introduce variations in the original pixel data, and consequently, applying diverse preprocessing methods can yield varying levels of accuracy in the collected data. This study explores the relationship between preprocessing techniques and the performance of deep learning models. In this paper, images are processed by both reducing and increasing their resolution and by applying 180° horizontal and vertical flips. Subsequently, the Edge Impulse platform (https://edgeimpulse.com/) is employed for image classification. The objective of this paper is to assess the impact of resolution adjustments and flipping on image classification accuracy through a series of repeated experimental analyses.