You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
我现在想基于这种无监督模型去获取异常样本的位置,然后将位置坐标作为提示点输入到SAM进行分割,这样获取异常缺陷的mask,相当于分割标签了,因为直接用无监督模型训练得到的mask似乎是不太准确的。我最近看到了RSPrompter: Learning to Prompt for Remote Sensing Instance Segmentation based on Visual Foundation Model这篇论文似乎表达了类似的意思
1、训练时所需要的fg_mask是在确定扩充异常样本的粘贴位置吗,避免缺陷粘贴在背景区域?
2、训练时分为两个阶段,分别是异常边界生成和边界引导,这两个训练过程是同步的吗?
3、选择的课件异常样本数量应该怎么选择呢,mvtec数据集默认每个类别选择10个异常样本,假设我有200个异常样本,3000个正常样本,异常样本数量选择多少呢?
The text was updated successfully, but these errors were encountered: