search

lmb-freiburg / flownet2

FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks
https://lmb.informatik.uni-freiburg.de/Publications/2017/IMKDB17/
Other
1k stars 318 forks source link

Augmenting during test #106

Closed adizhol closed 6 years ago

adizhol commented 6 years ago

Hi,

Thank you for publishing the code and weights. I have a small question - why is it necessary to augment input images when evaluating on test (production) images?

Thanks!

nikolausmayer commented 6 years ago

Hi, we are using the "augmentation" layer in test mode, but only for mean subtraction. It was just convenient for us to not switch to another layer for that. No geometry or color augmentation is done in production.

adizhol commented 6 years ago

Ok, thx

On Feb 12, 2018 11:51, "Nikolaus Mayer" notifications@github.com wrote:

Hi, we are using the "augmentation" layer in test mode, but only for mean subtraction. It was just convenient for us to not switch to another layer for that. No geometry or color augmentation is done in production.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/lmb-freiburg/flownet2/issues/106#issuecomment-364874056, or mute the thread https://github.com/notifications/unsubscribe-auth/AMeu9yj8vZ2H3w8wHPVyDWlTZCg1-g-Rks5tUAmWgaJpZM4SA52D .