Abstract: Convolution has been the core ingredient of modern neural networks, triggering the surge of deep learning in vision. In this work, we rethink the inherent principles of standard convolution ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results