Skip to content
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

您好,想请问一下Regression函数的实现 #12

Open
gaiyi7788 opened this issue May 18, 2022 · 4 comments
Open

您好,想请问一下Regression函数的实现 #12

gaiyi7788 opened this issue May 18, 2022 · 4 comments

Comments

@gaiyi7788
Copy link

gaiyi7788 commented May 18, 2022

您好,在您的实现中,Regression函数应该是同时集成了原文中的PFA和MDC模块。在阅读该函数时我注意到:

       self.stage1 = nn.Sequential(
            nn.Conv2d(256, 128, 3, padding=1, dilation=1),
            nn.BatchNorm2d(128),
            nn.ReLU(inplace=True)
        )
        self.stage2 = nn.Sequential(
            nn.Conv2d(256, 128, 3, padding=2, dilation=2),
            nn.BatchNorm2d(128),
            nn.ReLU(inplace=True)
        )
        self.stage3 = nn.Sequential(
            nn.Conv2d(256, 128, 3, padding=3, dilation=3),
            nn.BatchNorm2d(128),
            nn.ReLU(inplace=True)
        )

每个分支都是只使用了空洞卷积,而没有使用原论文中的1x1,3x3和5x5卷积,想问您这样实现的用意,比原文中加上这几个卷积效果会更好吗?谢谢!

@cjl09
Copy link

cjl09 commented May 18, 2022

他是按照原文实现的。仔细看,原文的意思是1X1,膨胀率=1的卷积而不是1x1后使用膨胀卷积。按照你的理解,后面的膨胀卷积他就没有卷积核大小了。那个图的确容易让人误会。

@gaiyi7788
Copy link
Author

他原文写的不是And each column consists of a single convolutional layer and a dilated convolutional layer,应该就是每个分支一个普通卷积一个空洞卷积吧,空洞卷积大小按图里的画的理解成3*3,我是这样觉得的

@wfs123456
Copy link
Owner

wfs123456 commented May 18, 2022 via email

@cjl09
Copy link

cjl09 commented May 18, 2022

实际上如何设置后面的空洞率,对结果的影响最多在1-2点内,每个人训练可能结果也不一样,不用太在意

---原始邮件--- 发件人: @.> 发送时间: 2022年5月18日(周三) 下午4:03 收件人: @.>; 抄送: @.>; 主题: Re: [wfs123456/CCTrans] 您好,想请问一下Regression函数的实现 (Issue #12) 他是按照原文实现的。仔细看,原文的意思是1X1,膨胀率=1的卷积而不是1x1后使用膨胀卷积。按照你的理解,后面的膨胀卷积他就没有卷积核大小了。那个图的确容易让人误会。 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.>

他问的不是如何设置后面的空洞率。应该是原文的意思是每列一个卷积再加一个空洞卷积。然后你的实现不是直接一个空洞卷积?我一开始也有这样的疑惑。多一层卷积有什么影响吗?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants