基于趋势驱动的弱监督增量语义分割
原文中文,约500字,阅读约需2分钟。发表于: 。Weakly Incremental Learning for Semantic Segmentation (WILSS) addresses the issue of conflicting predictions and catastrophic forgetting by proposing a tendency-driven relationship of mutual...
本文介绍了一种自适应的弱监督语义分割方法,通过划分图像内容为确定性和不确定性区域,提高准确性。实验证明该方法在PASCAL VOC 2012和MS COCO 2014上优于其他方法。