Scaling Up Computer Vision Neural Networks Using Fast Fourier Transform
This work addresses scaling issues in computer vision models for researchers and practitioners, but appears incremental as it applies existing FFT techniques to known bottlenecks.
The paper tackles the challenge of scaling computer vision neural networks by using Fast Fourier Transform to address limitations of large-kernel convolutions and Vision Transformers' quadratic complexity with high-resolution images, achieving unspecified improvements.
Deep Learning-based Computer Vision field has recently been trying to explore larger kernels for convolution to effectively scale up Convolutional Neural Networks. Simultaneously, new paradigm of models such as Vision Transformers find it difficult to scale up to larger higher resolution images due to their quadratic complexity in terms of input sequence. In this report, Fast Fourier Transform is utilised in various ways to provide some solutions to these issues.