Xinyi Ying

I come across the same problem. Is there any solutions?

根据[[TDAN]](https://github.com/YapengTian/TDAN-VSR-CVPR-2020)的机理，D3D应该也有对齐视频临近帧的作用，但是我们没有定量地探索过这个问题。 您如果需要继续研究D3D，可以使用一些toy examples做简单的可视化来验证“D3D是否有对齐视频临近帧的作用”这一问题。

Thanks for this comment. Please refer to ./dcn/test.py for example usage. The difference between DeformConv_d and DeformConvPack_d is also illustrated in ./dcn/test.py. Specifically, "DeformConvPack" is used for D3D deforming in...

Recently, deformable convolution is used to solve the video super-resolution problem. Specifically, Tian et al. [[R1]](https://arxiv.org/pdf/1812.02898.pdf) proposed a temporally deformable alignment network (TDAN) for video SR. In their method, the...

We can use D3Dnet in linux with gpu RTX3090 and the version of pytorch is 1.7.1 with cuda11.0.221 and cudnn8.0.5_0. The problem may be caused by the problem of "The...

Not yet. We don't know much about “colab notebook”. We are grateful if you can transfer our code to colab notebook as a contributor.

Sorry for reply late. Thanks for using our code. We build D3D on windows 10 so we have not met the same problems as you. Here we briefly introduce the...

@pearlie317 To compile D3D, "nvcc" and "cpp_extension" of pytorch 1.0.0 are needed. We do not find "gcc" in our PC. Our code is based on ''Deformable Convolution V2'' (https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch). You...

这段代码的目的是降低网络的参数量，即当在三维卷积的空域进行变形，只需要生成通道维度为2N的offset（其中N为卷积核的尺寸）。然而，在可变形三维卷积的底层C和cuda代码中，输入的offset的通道维数需要为3N。因此，需要将通道维度为2N的offset补零成通道维度为3N，再进行运算。 This code aims to reduce the number of parameters. That is, when only perform deformation in the spatial domain, only an offset with 2N channels is generated (N is...

是的，代码是基于[[DCNv2]](https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0)进行改进的。