Changelog

All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
[2.0.0] - 2019-10-07

Renamed max_points function to get_max_points to avoid confusion with a gotten max_points which is a good variable candidate name.
Refactored the geometry scaler to a class that uses .fit() and .transform() methods, analogous to the scikit-learn scaler objects.
More consistent names on geom_wkt parameter names, to signify passing in a well-known text representation of a geometry.
First backwards-incompatible release. This is necessary to resolve the loss of information from previous versions on polygon and multipolygon geometries with holes. Previous versions just ignored the holes to keep the encoding format identical to sketch-rnn but this project uses much simpler geometries. To include polygon holes without information loss, the format has been extended to include a length 2 one-hot vector for point classes of being part of a inner ring (interior ring or hole) or an outer ring (exterior). This format extension stretches the geometry vector format to shape(num_points, 7).
The padded output of simplified geometries, or geometry vectors extended to a fixed length are now padded with FULL_STOP bits. This is basically keep yelling to any neural net component to stop interpreting padding.
Included an optional is_last parameter to the vectorize_polygon method of the geovectorizer module. This allows you to specify whether a polygon being vectorized is the last in the (multi-part) geometry or not. Default is false.
First released pypi version on https://pypi.org/project/deep-geometry/: so basically everything