New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Adds customization options to the segmentation models and other improvements #263
Conversation
this is the newer version of fit_generator and should be used instead
…entationMapOnImage removed now unecessary nb_classes and replaced get_arr_int(), with get_arr()
just need to pass in the function in the cusomt_augmentation argument
This reverts commit d1d3935.
updates imageio and imgaug
(even though there should be no need to)
I believe that the Python 2.7 tests should be removed since Python 2 has been deprecated (January 1st 2020) and TensorFlow version 2.2 and up no longer support it. |
Yay all green!! @divamgupta I hope that this helps the community and it would be great if you checked out my changes! (this is my first pull request so any feedback would be greatly appreciated! ) |
these messages appear due to loading checkpoints but not using all the data
(Python 2 again...)
Hi @Marius-Juston thanks for the pull request. I will review it and get back to you. |
…th multi images test
added new folder and new test images for testing harder
in python 2.7 itertools.cycle seesm to give an infinite loop, so removed convoluted cycle and just use image seg pairs to get things to work
Just added some unit tests to verify everything works for the data generator functions |
in case the first layer was modified
Added the ability to change the number of image input channels to the models. |
@divamgupta I believe that I am done with my changes/feature additions. |
Okay now this was the last thing |
@Marius-Juston Thanks for the good work. I have merged the PR. If you want, you could also update the readme for the new features. |
Will do! |
Main additions:
Adds the ability to add custom callbacks to the model training sequences
Adds the ability to add custom augmentation functions to the model training
Adds the ability to add multi-image input with data augmentation to the models easily
Minor additions:
Uses TensorFlow model checkpoint instead of custom Keras one.
Uses tensorflow tf.train.latest_checkpoint function instead of custom find_latest_checkpoint function
Updates some of the deprecated augmentation function classes (using imgaug>=0.4.0)
For the auto-resume checkpoint boolean in training makes it so that it continues from the last checkpoint epoch number instead from 0 again
Uses the model.fit function instead of the model.fit_generator function
If the checkpoint base folder does not exist, create it
Updates the dataset visualization part of the library to accommodate different augmentation functions and image size