Shifting Color Space for Image Classification using Genetic Programming
David Herrera-Sánchez, Héctor‐Gabriel Acosta‐Mesa, Efrén Mezura‐Montes, Aldo Márquez-Grajales · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2024
Image classification is an important application in different areas of computer vision. Genetic Programming (GP) has demonstrated an excellent capability to deal with this task. However, many approaches transform images to grayscale due to the computational cost of working with color spaces. Therefore, possible significant image features for classification could be discarded early. Consequently, our proposal incorporates features extracted from different color spaces for image classification through GP in three datasets. Furthermore, the GP's flexibility allows the images' color space conversion during the evolutionary process. According to the dataset, the final solution determines whether a specific color space was used or if it was necessary to change to another to extract meaningful features from the images. The results demonstrate that changing the color space in some data sets was necessary to achieve competitive classification accuracy.