Coordinates

Coordinate convention handling for converting between annotation, numpy, and shapely coordinate systems.

nimbusimage.coordinates

Coordinate convention handling for NimbusImage annotations.

Three coordinate systems: 1. Annotation: {'x': pixel_x, 'y': pixel_y} — x horizontal, y vertical 2. Numpy: array[row, col] = array[y, x] 3. Shapely: Point(x, y) — x horizontal, y vertical

Convention in this package: - Shapely x = annotation x = image column (horizontal) - Shapely y = annotation y = image row (vertical) - No x/y swap for shapely conversions - For numpy masks: row = annotation y, col = annotation x - 0.5 offset: annotation coords are at pixel top-left corners; scikit-image draws at pixel centers

annotation_to_polygon(coordinates)

Convert annotation coordinate dicts to a shapely Polygon.

No x/y swap — annotation x maps to shapely x (horizontal).

annotation_to_point(coordinates)

Convert annotation coordinate dicts to a shapely Point.

For single-coordinate annotations, returns that point. For multi-coordinate (polygons), returns the centroid.

polygon_to_coordinates(polygon)

Convert a shapely Polygon to annotation coordinate dicts.

Strips the duplicate closing point that shapely adds.

point_to_coordinates(point)

Convert a shapely Point to annotation coordinate dicts.

coordinates_to_mask(coordinates, shape)

Convert annotation coordinates to a boolean mask.

Applies the 0.5 offset: annotation coords are at pixel top-left corners, but rasterization uses pixel centers.

Parameters:

Name	Type	Description	Default
coordinates	list[dict]	List of {'x': ..., 'y': ...} dicts.	required
shape	tuple[int, int]	(height, width) of the output mask.	required

Returns:

Type	Description
ndarray	Boolean numpy array of the given shape.

mask_to_coordinates(mask)

Convert a boolean mask to annotation coordinates.

Uses contour finding and adds the 0.5 offset back.

Returns:

Type	Description
list[dict]	List of {'x': ..., 'y': ...} dicts forming the polygon boundary.