search

szagoruyko / cvpr15deepcompare

Code and models for "Learning to Compare Image Patches via Convolutional Neural Networks"
http://imagine.enpc.fr/~zagoruys/deepcompare.html
Other
468 stars 169 forks source link

Interpreting Network Outputs #13

Closed Supersak80 closed 8 years ago

Supersak80 commented 8 years ago

Hi @szagoruyko , when I run test patches through a trained network, how am I to interpret the output similarity scores? It doesn't appear that they are normalized to [0,1].

Supersak80 commented 8 years ago

@szagoruyko Sergey, your input is greatly appreciated with the question above.

szagoruyko commented 8 years ago

@Supersak80 sorry, I forgot to answer. the outputs are not normalized to [0,1], higher the score - more similar patches are.

Supersak80 commented 8 years ago

@szagoruyko Thank you, Sergey!