Scaling Up Computer Vision Neural Networks Using Fast Fourier Transform

While Fourier Transform has seen many applications in data compression and signal processing, it's applications in Deep Neural Networks are limited. They are often only used for Medical Imaging based applications. Here, I discuss three approaches of scaling up Neural Networks for computer vision using Fast Fourier Transform (FFT). Note: python libraries were used for FFT as they provide extremely efficient cuda-based approaches for FFT on the GPU. The report only contains part of the code; the entire code-base is very large and includes the dataloaders, hyperparameter configurations, scripts to test fps, datasets, training and validation engines, and the model implementations themselves.