conv_arithmetic
conv_arithmetic copied to clipboard
vdumoulin
A technical report on convolution arithmetic in the context of deep learning
In PyTorch, when the stride is 0, RuntimeError: non-positive stride is not supported will be raised. Meanwhile, the default argument stride of nn.Conv2d is 1. https://pytorch.org/docs/stable/generated/torch.nn.Conv2d.html#:~:text=%2C%20kernel_size%2C-,stride%3D1,-%2C%20padding%3D
I believe from the pdf that the cyan should be the input in the transposed convolution gifs, and the blue should be the output. It should read: N.B.: Blue maps...
* Support for non-square kernel-size/stride/padding/dilation * Automatic computation of output size from the other dimensions (if not explicitly set) * Support for custom colors * Ability to generate all steps...
I am attaching the two figures taken from the [arXiV](https://arxiv.org/pdf/1603.07285v2.pdf). In convolution: it says the padding is 2x2 and in transposed, it also says the padding is 2x2. Shouldn't it...
Thanks guys for this amazing report, it helped me a lot in understanding convolutional arithmetic. I think I found a slight error in the description of the figure 4.7 description...
what about make a Deformable Convolutional's gif image,thanks!
Aren't there mistakes in the images 1 to 8 for numerical_padding_strides-*.pdf ? For instance shouldn't the second element of the first row be 5 ? Here is what is computed:...
Part of the figure 4.6 caption states: > It is equivalent to convolving a 3 × 3 kernel over a 2 × 2 input I think that should be: *It...
