A highly customisable generator for parametrically varied 2D shapes. Useful for psychological experiments and proof-of-concept Machine Learning simulations.\\ The generator returns single .png files for each generated image, as well as a MATLAB .mat file in which all images are stored as uint8 matrices, together with a copy of all parameters and labels for each image.
If you run the script
python main.py
a few demo shapes with default settings will be generated and stored in your project folder as .png files.
All settings are defined in the main.py file via argparse flags. Therefore, the user can easily specify parameters as command line arguments without having to mess with python code.
python main.py --shapes ellipse && python main.py --shapes ellipse --stim_ellipse_ratio 0.3 1.2
python main.py --shapes star5 star8 star18
python main.py --shapes poly3 poly5 poly7
python main.py --shapes star5 --to_transform colour --num_transformations 20
creates 20 stars with fie vertices, parametrically varied in colour from red to blue\\
python main.py --shapes rect --to_transform scale rota --num_transformations 5
creates a 5 by 5 grid of rectangles, with parametrically varied scale and rotation
The following parameters can be changed from the command line to customise the stimuli
--outdir output directory
--name file name
--canvas_size size of image [px]
--canvas_bccol background colour [r g b]
--num_stimuli number of stimuli per shape (duplicates)
--num_transformations number of linearly spaced transformations
--shapes shapes to generate
--to_transform dimensions to transform
--stim_poly_size default size of polygon
--stim_star_size default size of star
--stim_rect_ratio default width to height for rectangles
--stim_ellipse_ratio default width to height for ellipses
--stim_scale default width and height scale
--stim_rota default rotation
--stim_trx default translation along x
--stim_try default translation along y
--stim_colour default colour [r g b]
--rng_scale range of scale transformation
--rng_trx range of translation along x
--rng_try range of translation along y
--rng_rota range of rotation transformation