Closed ivanzvonkov closed 3 years ago
@ivanzvonkov thanks for this!
Very nice library guys and super well documented - we are very happy to see Kornia in here :)
Regarding the current issue - why not passing scale = torch.ones(b, 2)
?
https://github.com/greentfrapp/lucent/blob/master/lucent/optvis/transform.py#L71
An alternative is to use kornia.rotate
which doesn't take into account the scale as input.
Besides, I detected few other functionalities in your code that could be used directly with kornia. Let me list them:
kornia.deg2rad
https://github.com/greentfrapp/lucent/blob/master/lucent/optvis/transform.py#L69kornia.tensor_to_image
https://github.com/greentfrapp/lucent/blob/master/lucent/optvis/render.py#L132kornia.normalize
https://github.com/greentfrapp/lucent/blob/master/lucent/optvis/transform.py#L106Last, what's the intuition or if you have any reference for _linear_decorrelate_color
?
I believe with a proper API design would be great to have it in kornia.
kornia 0.4.1 released recently and it includes a breaking change to
get_rotation_matrix2d
I've opened an Issue to hopefully address this soon: https://github.com/kornia/kornia/issues/742 but in the meantime, I suggest freezing kornia at 0.4.0 so that the
random_rotate
transform continues working as before