New makedatasets - Githubissues

Doodleverse / segmentation_gym

A neural gym for training deep learning models to carry out geoscientific image segmentation. Works best with labels generated using https://github.com/Doodleverse/dash_doodler

MIT License

45 stars 11 forks source link

New makedatasets #99

Closed dbuscombe-usgs closed 2 years ago

dbuscombe-usgs commented 2 years ago

minor bug fixes in seg_images_in_folder
added print statements to seg_images_in_folder as per https://github.com/Doodleverse/segmentation_gym/issues/96
some comment cleanup in do_train and make_datasets
updated vanilla unet config file to reflect new config parameters
updated yml file, see https://github.com/Doodleverse/segmentation_gym/issues/93
made changes to some utilities, and moved gen_saved_model to utils, see https://github.com/Doodleverse/segmentation_gym/issues/94

dbuscombe-usgs commented 2 years ago

I think once this is reviewed and merged, it is time for a 'release'. Coinciding with the paper.. I think we have now gotten Gym to the point where is "stable" (jeez, famous last words!)

I'll then tag the release in the new zenodo release for the test dataset