The Effect of Preprocessing with Gabor Filters on Image Classification Using CNNs
Akito Morita, Hirotsugu Okuno · Proceedings of International Conference on Artificial Life and Robotics · 2022
In image classification tasks, preprocessing of input images is one of the promising approaches for improving the performance.In this study, we investigated the effect of neuro-inspired preprocessing, such as Gabor filtering.We compared the averaged classification accuracy of multiple CNNs with the following three types of preprocessing: no preprocessing, Gabor filtering, and calculation of the difference between two Gabor filtered signals in the opposite color channels.The results showed that Gabor filtering increased the classification accuracy.