dcurro/cannyedgepytorch

cannyedgepytorch's Issues

batch issue

I think there is something wrong in the forward function of networks as it doesn't supports arbitary batch sizes, it only supports batch size of 1

License requested

It would be wonderful if there was a license associated with this repo. I'd love to use it in my research, but I can't without knowing its license; public-use permissions are not implied by the absence of a license in a GitHub repo.

Backward anomaly problem

Thank you very much for your work, I found an anomaly when I try to differentiate it and do the backward step, your Net has many in-place operations which makes it impossible to calculate the gradients. If anyone encounters the same problem replace the first few lines in the forward function to:

    def forward(self, img):

        blurred_img_r = self.gaussian_filter_vertical(self.gaussian_filter_horizontal(img[:, 0:1]))
        blurred_img_g = self.gaussian_filter_vertical(self.gaussian_filter_horizontal(img[:, 1:2]))
        blurred_img_b = self.gaussian_filter_vertical(self.gaussian_filter_horizontal(img[:, 2:3]))

        grad_x_r = self.sobel_filter_horizontal(blurred_img_r)
        grad_y_r = self.sobel_filter_vertical(blurred_img_r)
        grad_x_g = self.sobel_filter_horizontal(blurred_img_g)
        grad_y_g = self.sobel_filter_vertical(blurred_img_g)
        grad_x_b = self.sobel_filter_horizontal(blurred_img_b)
        grad_y_b = self.sobel_filter_vertical(blurred_img_b)

runtime error

Hello, thanks for your great work!

But there are some errors when I try to use your code.
Some runtime errors occur in the forward part, line 120
channel_select_filtered_positive = all_filtered.view(-1)[indices.long()].view(1,height,width)
RuntimeError: index -9223372036854775808 is out of bounds for dimension 0 with size 524288

I am a new hand in coding and I can not find the reasons. Can you help me to see why it occurs? Thanks a lot!

Two bugs encountered

Great work - thanks a lot for this!

When using the code on some test images, I had to adjust the following pieces of code:

In Net.forward() after line 135 I added the statement

is_max = torch.unsqueeze(is_max, dim=0)

In Net.__init__() change line 79 to prevent a dimension mismatch as follows:

self.directional_filter.weight.data.copy_(torch.from_numpy(all_filters[:, None, ...]))