ImageBoundaryPointsGeneration takes an image..

.. identifies the " region " in it and returns a list of points bounding the region.

This is meant to be used with a procedural generation project that generates 2D tile maps from provided tile sets within specified regions. It facilitates specifying those regions by uploading pictures in which the specification is clearly present.

Example:

The above is a region assembled from pieces I picked up at a fair.

Approximation is done by calling RegionBordersApproximator.approximate:

RegionBordersApproximator.approximate(base_image_data=base_image_data, base_image_width=400, base_image_height=300, min_canny_threshold=100, max_canny_threshold=200, min_segmenting_threshold=70, max_segmenting_threshold=255, contour_thickness=2)

• base_image_data: image in Base64
• base_image_width, base_image_height: image width/height
• min_canny_threshold, max_canny_threshold: Canny edge detection thresholds
• min_segmenting_threshold, max_segmenting_threshold: Image segmenting thresholds
• contour_thickness: thickness of lines drawn for region borders

Results are output as

@dataclass(frozen=True)
    class RegionBordersApproximation:
        border_points: dict[uuid4, Point]
        border_segments: tuple[RegionBordersApproximator.RegionBorderSegment]

RegionBorderSegment.point_Id values correspond to border_points keys.

Examples (visual):

Borders highlighted:

Borders & boundary points highlit:

The idea is that the border points can be more finely, manually tweaked from this initial state. Future iterations will reliably automate that tweaking.