Support non-RGB images with --img_channels option #14
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The expected number of image channels can now be expressed through the "--img_channels" option. This option changes the generator and discriminator architectures to generate/expect the given channel count and changes the data loading mechanism to expect and - if necessary - convert images to have this number of channels. - Fixes the issue with occasional grayscale images in RGB photo data sets reported in akanimax#5 because if --img_channels=3 (default), grayscale images are automatically converted to RGB in preprocessing. - Adds support for training with grayscale images, as requested in akanimax#5. - Makes it easier to implement training with multi-channel images that have more than 3 channels (which I'm planning to try soon). - Enables training with RGB data sets in grayscale mode by simply setting --img_channels=1.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The expected number of image channels can now be expressed through the
--img_channelsoption. This option changes the generator anddiscriminator architectures to generate/expect the given channel count
and changes the data loading mechanism to expect and - if necessary -
convert images to have this number of channels.
reported in I have met this error when run train.py ... #5 because if
--img_channels=3(default), grayscale imagesare automatically converted to RGB in preprocessing.
have more than 3 channels (which I'm planning to try soon).
--img_channels=1.Disclaimer: I have not tested my changes with the usual data sets for image synthesis. I have only tried it with two small toy data sets: one of RGB images and one of grayscale images. My test trainings run without errors and the generated data visualization works, but my trainings don't converge yet and probably need some hyperparameter tuning (the discriminator loss is at 0 very often).