Super-resolution image by decomposing the 2D convolution of LKA into 1-D kernels with LCAN
Super-resolution image by decomposing the 2D convolution of LKA into 1-D kernels with LCAN
Large coordinate kernel attention network for lightweight image super-resolution
arXiv paper abstract https://arxiv.org/abs/2405.09353
arXiv PDF paper https://arxiv.org/pdf/2405.09353
The multi-scale receptive field and large kernel attention (LKA) … improve performance in … image super-resolution task … propose the multi-scale blueprint separable convolutions (MBSConv) as … building block with multi-scale receptive field, it … focus on … learning … multi-scale information
… revisit the key properties of LKA in which … find that the adjacent direct interaction of local information and long-distance dependencies is crucial to provide … performance.
… propose a large coordinate kernel attention (LCKA) module which decomposes the 2D convolutional kernels of the depth-wise convolutional layers in LKA into horizontal and vertical 1-D kernels.
LCKA enables the adjacent direct interaction of local information and long-distance dependencies not only in the horizontal direction but also in the vertical.
… LCKA allows … large kernels in the depth-wise convolutional layers to capture … contextual information, which … improve the reconstruction … and … lower computational … and memory …
Integrating MBSConv and LCKA, … propose a large coordinate kernel attention network (LCAN) … LCAN with the lowest model complexity achieves superior performance …
Stay up to date. Subscribe to my posts https://morrislee1234.wixsite.com/website/contact
Web site with my other posts by category https://morrislee1234.wixsite.com/website