Closed Kelly-ZH closed 3 years ago
@Kelly-ZH
y
in the line is W in section 4.1 of MPIIFaceGaze paper.@Kelly-ZH
- Yes. And now I realized that this comment about the size of ResNet was wrong. Sorry about that. I fixed it.
- The
y
in the line is W in section 4.1 of MPIIFaceGaze paper.
Thank you for your reply. About quetion 2, the paper propose to use additional three convolutional layers to learn spatial weights, and you only use one 1*1 convolutional layer, is it?
@Kelly-ZH
For ResNet, yes.
Hi, hysts. I drew the model of resnet_14 and resnet_preact_8, you can refer to the attachments for details. Could you please help me to see what's wrong with my work, thank you very much.
Yours, Kelly
@Kelly-ZH
I'm confused. Obviously, nothing is attached to your reply... Did you forget to drag and drop your images or something? You can check if the image can be displayed correctly by clicking on the preview tab above before hitting the comment button.
I am sorry, I didn’t upload the attachment successfully. I upload them as pictures again.
发件人: hysts 发送时间: 2021年4月9日 1:53 收件人: hysts/pytorch_mpiigaze 抄送: Kelly-ZH; Mention 主题: Re: [hysts/pytorch_mpiigaze] Questions about resnet_14 (#41)
@Kelly-ZH I'm confused. Obviously, nothing is attached to your reply... Did you forget to drag and drop your images or something? You can check if the image can be displayed correctly by clicking on the preview tab above before hitting the comment button. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.
@Kelly-ZH
Are you replying to an email from GitHub? I'm not sure what's going on, but anyway, I can't see any images in this issue comment.
@Kelly-ZH
Yes, now I can see them. And they look fine.
@Kelly-ZH
Yes, now I can see them. And they look fine.
Thank you. For some personal reasons, I will delete the picture. Hope you can understand.
Hi , I have two other questions. Can you answer it for me? Thank you.
Thank you.
@Kelly-ZH
For ResNet, yes. Hi, hysts the paper propose to use additional three convolutional layers to learn spatial weights, and why did you only use one 1*1 convolutional layer? Could you answer it for me?
Thank you very much.
Hi, hysts. I studied the code further and encountered new problems, and described as bellow:
Thank you very much. I am looking forward your reply.
Yours, Kelly