create_non_correspondences

[JoeyChia]

I have found that the following code may be not right.

pytorch-dense-correspondence/dense_correspondence/correspondence_tools/correspondence_finder.py

Line 269 in c3b068a

 ones = torch.zeros_like(diffs_0_flattened) 

Should it be "ones = torch.ones_like (diffs_0_flattened)" ? And I also interesting in why using line 277 ~280 to generate random_vector ?

[Pete Florence]

Hi, I think you are right, the code is not doing what is intended there. The purpose of that snippet was to change the randomly sampled non match if it is too close to the true match. In the end it will not matter too much since it is a low probability event to randomly sample the true match (1 out of 307,200). I can look more at it later and see if it makes a difference to fix it. Thanks!

…

On Wed, Nov 28, 2018 at 2:06 AM HugeKangaroo ***@***.***> wrote: I have found that the following code may be not right. https://github.com/RobotLocomotion/pytorch-dense-correspondence/blob/c3b068adca006b828248fdf16b00aa7603d462e2/dense_correspondence/correspondence_tools/correspondence_finder.py#L269 Should it be "ones = torch.ones_like (diffs_0_flattened)" ? And I also interesting in why using line 277 ~280 to generate random_vector ? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#182>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AFYQqNi5-hk76o9xeYKgoH8ut6Cx7gWWks5uzjYGgaJpZM4Y2_Pb> .