Correcting Patch Importance for Dominant Patch Identification
We multiplied the PageRank values of each patch by their respective logical labels (as weight values of 0 and 1) to correct for the importance of patches, which helped the model identify dominant patches.
原文地址: https://www.cveoy.top/t/topic/kJ1 著作权归作者所有。请勿转载和采集!