Varied Image Data Augmentation Methods for Building Ensemble

Convolutional Neural Networks (CNNs) are used in many domains but the requirement of large datasets for robust training sessions and no overfitting makes them hard to apply medical fields similar fields. However, when quantities samples cannot be easily collected, various methods can still applied stem problem depending on sample type. Data augmentation, rather than other methods, has recently been under spotlight mostly because simplicity effectiveness some more adopted methods. The research question addressed this work is whether data augmentation techniques help developing efficient machine learning systems different classification purposes. To do that, we introduce new image that make use like Fourier Transform (FT), Discrete Cosine (DCT), Radon (RT), Hilbert (HT), Singular Value Decomposition (SVD), Local Laplacian Filters (LLF) Hampel filter (HF). We define ensemble by combining classical with newer ones presented here. performed an extensive empirical evaluation 15 validate our proposal. obtained results show newly proposed very effective even alone. ensembles trained augmentations outperform best approaches reported literature as well compete state-of-the-art custom All resources available at https://github.com/LorisNanni .