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

Add deformable convolution DCNv2 support #1788

Closed
breznak opened this issue Jan 24, 2020 · 2 comments · Fixed by #2791
Closed

Add deformable convolution DCNv2 support #1788

breznak opened this issue Jan 24, 2020 · 2 comments · Fixed by #2791

Comments

@breznak
Copy link

breznak commented Jan 24, 2020

Recently there was a great PR 👍 adding deformable convolution support DCN (v1) support to this repo. Meanwhile improved version DCNv2 appeared, so I'd like to support that.

@ksnzh
Copy link

ksnzh commented Aug 28, 2020

Mmdetection has two implementations: 1) deform_conv 2) modulated_deform_conv. The difference between those two is that modulated_deform_conv one more argument mask. Also, they have different implementations in cpu and cuda, and he torchvision's version is the former.

class DeformConvFunction(Function):

    @staticmethod
    def forward(ctx,
                input,
                offset,
                weight,
                stride=1,
                padding=0,
                dilation=1,
                groups=1,
                deformable_groups=1,
                im2col_step=64):

class ModulatedDeformConvFunction(Function):

    @staticmethod
    def forward(ctx,
                input,
                offset,
                mask,
                weight,
                bias=None,
                stride=1,
                padding=0,
                dilation=1,
                groups=1,
                deformable_groups=1):

@inspiros
Copy link

inspiros commented Sep 2, 2020

In the mean time, my Covid work might help someone https://github.com/inspiros/tvdcn

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

Successfully merging a pull request may close this issue.

3 participants