linyuxi_loretta的博客convolutional layer就是,里面有一排filter,每个都是一个3 x 3 x channel 这么大的tensor,每个filter的作用就是抓取图片里的某个pattern,每个pattern需要在3 x 3 x channel那么小的范围内,才能被filter抓出来。...
Pengsen Ma的博客Abstract Video inpainting aims to fill the given spatiotemporal holes with realistic appearance but is still a ... Recent works introduce the promising Transformer architecture into deep video i