Hull Cover Conditioned Unit Disk Graph

⠀⠀

A generator for unit disk graphs conditioned on hull cover. The model first generates uniformly distributed points in 2 dimensions. Based on the input image points which are not covered with black (conditioning) are deleted from the initial set of points. Using the remaining point a fixed r-radius unit disk graph is grown. Finally, the edge list of the graph is saved with a large resolution plot of the graph.

The code takes an input black and white jpeg file. Every pixel indicates whether the data point generated randomly is covered by black in the image or not. Points without a cover are dropped. The /input/ folder contains a Mickey Mouse head, a T-rex and a flower as an example input image.

Options

Learning of the embedding is handled by the src/main.py script which provides the following command line arguments.

Input and output options

  --input-path    STR     Input image path.         Default is `input/trex.jpg`.
  --output-image  STR     Output image path.        Default is `output/edges/trex_edges.csv`.
  --output-edges  STR     Edge list path.           Default is `output/plot/trex_graph.png`.

Model options

  --node-color     STR         Node coloring in plot.         Default is `green`. 
  --point-number   INT         Number of points in space.     Default is 10,000.
  --dpi            INT         DPI of plot.                   Default is 500.
  --radius         FLOAT       Radius for graph generation.   Default is 0.04.
  --line-width     FLOAT       Line width.                    Default is 0.2.
  --alpha          FLOAT       Alpha for color strength.      Default is 0.3.
  --node-size      FLOAT       Node size in plot.             Default is 12.0.

Examples

The following commands learn a graph embedding and write the embedding to disk. The node representations are ordered by the ID.

Creating a random graph from the default T-rex image.

python src/main.py

Creating an random graph from the flower image. The graph generation and plotting parameters are changed to create a high quality result.

python src/main.py --input-path input/flower.jpeg --output-image output/plot/flower.png --output-edges output/edges/flower.edges --node-color "red" --radius 0.07 --alpha 0.5

License

• GNU