像素级图像细化算法PAMR
时间: 2023-10-22 11:30:11 浏览: 42
像素级图像细化算法PAMR是一种通过多次迭代的方式对粗糙的Mask进行细化的方法。它利用Mask在原始图像上对应像素点邻域的特性来更新Mask像素点的标签。作者定义了像素级亲和力(pixel-level affinity),来反应图像上某一像素点与其邻域上的像素点之间的亲和度。PAMR模块的参数在训练时不参与优化过程。通过对不同膨胀率的3×3卷积核的组合,PAMR模块可以提取像素点的邻域信息。然后,PAMR模块产生的细化mask将对网络输出的粗糙mask进行自监督,使用交叉熵损失进行训练。<span class="em">1</span><span class="em">2</span><span class="em">3</span>
#### 引用[.reference_title]
- *1* [Single-Stage Semantic Segmentation from Image Labels 笔记【CVPR 2020】](https://blog.csdn.net/weixin_41321482/article/details/108929948)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v93^chatsearchT3_2"}}] [.reference_item style="max-width: 50%"]
- *2* *3* [CVPR 2022 | 武大&京东提出:使用Transformer的端到端弱监督语义分割](https://blog.csdn.net/amusi1994/article/details/124811910)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v93^chatsearchT3_2"}}] [.reference_item style="max-width: 50%"]
[ .reference_list ]