fenghao195

Results 7 issues of fenghao195

您好,按照setting1中设置训练50,0000。在每隔2000个迭代会进行一次验证,这些验证保存的结果让我满脸问号???以set5中的‘baby’为例,下面是每次验证保存的结果分别是lr,2000iter,4000iter,6000iter,8000iter,10000iter。10000iter后的图片和10000iter结果差不多,都是黑屏: 原始lr: ![sig1 8_baby_LR](https://user-images.githubusercontent.com/76979950/178260961-fc8e6672-0ae1-45a5-8841-35055229ce2a.png) 2000iter: ![sig1 8_baby_2000](https://user-images.githubusercontent.com/76979950/178261447-36b3e3f8-3f0a-44b7-9040-24af91792ef7.png) 4000iter: ![sig1 8_baby_4000](https://user-images.githubusercontent.com/76979950/178261501-230171ea-41eb-4a36-b6f2-c12b958da382.png) 6000iter: ![sig1 8_baby_6000](https://user-images.githubusercontent.com/76979950/178261534-0b5b95e5-a7b9-49ec-b937-75bd9ac241f5.png) 8000iter: ![sig1 8_baby_8000](https://user-images.githubusercontent.com/76979950/178261588-89c3dfe6-6c40-4608-a0d8-562065b6742a.png) 10000iter: ![sig1 8_baby_10000](https://user-images.githubusercontent.com/76979950/178261619-8f27d512-4859-4c9d-b71a-87e8fce8ca1f.png) 后面从第10000iter到20,0000的结果都是全黑图片。这是在第8000iter后过拟合了吗。第4000iter和6000iter结果也很离谱。求救,您认为为什么会这样

您好,注意到您合成数据是用的各向同性高斯模糊核,并没有实现各向异性高斯模糊核的代码。你是认为各向同性高斯模糊核更有一般性吗?

在打印的网络中并没有看到动态生成的weight和bias,这样是可以训练的嘛?(ps:此前在训练层注意力中,也用到了beta=nn.parameter,但是定义的参数beta值一直都没有改动!!!)。能问下您的pytorch版本嘛?

Hi,I find that self.gamma is not trained,and requires_grad of tensor is True.model do not contain it

![image](https://user-images.githubusercontent.com/76979950/193233250-48c42ef9-1d91-4e43-84b0-01d51166022f.png) 您好,这是condconv中的bias,在训练中,它的值一直没有改变。之前在训练层注意力时候,就有这个问题,您知道这是为什么嘛?

Good evening, boss! I recently discovered your work about MANet.I found that the length of the gaussian kernel your method generated is equal to 18.Does this setting have any specific...

您好,大佬。我们在做对比试验中发现DCLS生成的模糊核偏离真实核,这应该是您重构后的核,应该怎么将其恢复呢? ![img_043](https://user-images.githubusercontent.com/76979950/233354982-bbea2809-30a8-4ae3-aaec-840acbbc4ba4.png) ![img_096](https://user-images.githubusercontent.com/76979950/233355013-053a687c-80dd-4c59-9901-7ff9f72ca384.png) DCLS生成的核 ![img_043](https://user-images.githubusercontent.com/76979950/233355052-435f9792-9de7-4d89-b50c-a224f81bebf8.png) ![img_096](https://user-images.githubusercontent.com/76979950/233355075-5fc43c15-29c1-4bac-9865-6625fbaad2f3.png) GT核