PeLK: Parameter-efficient Large Kernel ConvNets with Peripheral Convolution

arXiv:2403.07589v2 Announce Type: replace
Abstract: Recently, some large kernel convnets strike back with appealing performance and efficiency. However, given the square complexity of convolution, scaling up kernels can bring about an enormous amount of parameters and the proliferated parameters can induce severe optimization problem. Due to these issues, current CNNs compromise to scale up to 51x51 in the form of stripe convolution (i.e., 51x5 + 5x51) and start to saturate as the kernel size continues growing. In this paper, we …

abstract arxiv cnns complexity convolution cs.cv current efficiency however kernel optimization parameters performance scaling scaling up square strike type