-
Notifications
You must be signed in to change notification settings - Fork 27
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
论文中的1000样本对/类异常 #60
Comments
我也遇到这个问题,请问您解决了嘛 |
请问是否能提供下测试出来的指标数据,这504个数据是能够大致获得最终的指标的,可能具体某些类会有1-2个点的浮动 |
@duanyuanyi 我在 MVTec 课程中得到了类似的数字。另外,@sjtuplayer 你能告诉我为什么上一代的图像和当前几代的图像有巨大的差异吗?此外,您在论文中说您使用了 1000 张图像来进行下游分类、定位和拘留任务。这与你的主张相矛盾。 |
@sjtuplayer 看来您还执行了两次推理,查看榛子文件夹,因为从技术上讲,“ori”文件夹中应该有好的图像,但我们观察到缺陷图像的可见性较低。我不明白为什么你不断更改生成的数据集并且你说有多个版本的数据。这意味着您不知道您根据哪些数据计算结果。太棒了! |
Hi. When generating the data, there are multiple hyperparameters we can modify (e.g., cfg, employing mask generation models in different training steps), and each may have an influence on the results. So, there are different versions of data. But they are all generated by the same anomaly generation model. Previous data for hazelnuts is mistakenly placed, So, I replaced them which can better reproduce the reported AP score. But the data in the 'ori' directory still have some problems. I will fix it soon. |
使用您开源的504样本对/类异常在异常检测任务上没有达到论文中的效果,您能否提供论文中的生成数据集呢,我们将在论文中引用您的工作。
The text was updated successfully, but these errors were encountered: