gasoonjia / tvnet_pytorch Goto Github PK

Hi,

I am interested in making derivative works of this code, but I am unable to as there is no license specified. Can you upload a LICENSE.md, or add something to readme specifying which open source license this is released under?

Related code here

It seems that if we change to use F.grid_sample() from spatial_transform(), the third param that specifies the output shape should be removed. Since F.grid_sample() does not accept this param at all.

Thanks for your great code!
self.centered_gradient_kernels = self.get_centered_gradient_kernel().train(False)
When I remove the 'train(False)', I found that the paramters of get_centered_gradient_kernel cannot be trained and it retains unchanged.I guess that maybe some ops are non-differentiable. And as the paper said "We also propose to relax the convolutional filters in Eq. (7)-(9). The original convolutions are used to derive the (numerical) gradients and divergences", I wonder why this implement and the original implement don't set this paramter trainable. If so, the trainable paramter is only u0. And the result of trained model is similar to the untrained model's.

Hi. Thank you for your code!
I want to train this network for tracking application, and when I change batch_size parameter it doesn't affect the data. Is this behavior expected? is it a bug?
thanks