geo-deep-learning
geo-deep-learning copied to clipboard
Allow float in min_annotated_percent
Currently min_annotated_percent is required to be an integer, this appears an arbitrary constraint. Plus the background pixel percentage computed at https://github.com/NRCan/geo-deep-learning/blob/8d85078a1aeae791244807660c5cd5119cca7d83/images_to_samples.py#L239 is rounded with 1 decimal, why is that?