This repository has been archived by the owner on Mar 17, 2021. It is now read-only.
Histogram normalisation fails on RGB images #262
Comments
tomvars
pushed a commit
to tomvars/NiftyNet
that referenced
this issue
Oct 27, 2018
Additional arguments are added when calling `self.net`
in classification, segmentation, regression applications
to facilitate `keep_prob` network argument.
Network's should be modified to have the
`layer_op(a, b, c, **unused_kwargs)` signature
to tolerate the irrelevant arguments.
6 tasks
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
As of dev-branch commit a0eab6e:
The file reader correctly recognises the channels of an RGB image as distinct modalities. However, when enabling histogram normalisation via "normalisation = True" in the network section of the model configuration file, an exception is thrown because the histogram normalisation code incorrectly expects unimodal data.
Apologies, the exception stacktrace could not be attached, but to reproduce: load any RGB bitmap data set and set "normalisation = True" in the config file.
The text was updated successfully, but these errors were encountered: