Abstract: Pruning convolutional kernels is a crucial method for achieving model lightweighting. However, current APoZ (Average Percentage of Zeros) based pruning algorithms often overlook the ...