Shape generate

The file shape_generate.py has multiple functionality to create synthetic dataset of symmetrical objects. The methods creates two folders, namely, images and masks. The images are constructed from the Gaussain distribution.

Installation

This repo is managed with poetry. To try out this package, do:

$ poetry install
$ poetry run symmsyn --help

Examples

The following examples show the method to create the dataset

1. Create a dataset of 200 images of size=256 x 256 with circle objects of radius=10 and density=20

poetry run symmsyn --job_type "single" --n_image 200 --folder_name "circle_data" --object_type "circle" --image_dims 256 --circle_radius 10 --start_count 45 --end_count 50

2. Create a dataset of 100 images of size=512 x 512 with square objects of side=10, density=35

poetry run symmsyn --job_type "single" --n_image 100 --folder_name "square_data" --object_type "square" --image_dims 512 --start_count 34 --end_count 35 --square_size 10 10

3. Create a dataset of 100 images of size=512 x 512 with triangle objects of varying size, density=35

poetry run symmsyn --job_type "single" --n_image 100 --folder_name "triangle_data" --object_type "triangle" --image_dims 512 --start_count 34 --end_count 35 --size_vary True

4. Create a dataset of 500 images of size=256 x 256 with circle+ellipse objects of radius=10, minor_radius=5, major_radius=10, circle density=35, ellipse density=10

The elliptical objects will be equally divided into horizontal and angular objects. Specify angle parameter if needed.

poetry run symmsyn --job_type "multi" --n_image 100 --folder_name "circle_ellipse_data" --image_dims 256 --circle_radius 10 --circle_start 34 --circle_end 35 --add_ellipse True --ell_count 10 --m_radius 10 --n_radius 5 --ell_angle 45

5. Create a dataset of 500 images of size=256 x 256 with circle+ellipse+triangle objects of varying sizes and circle density=10, ellipse density=10, triangle density=40

The elliptical objects will be equally divided into horizontal and angular objects. Specify angle parameter if needed.

poetry run symmsyn --job_type "multi" --n_image 500 --folder_name "circle_ellipse_tri_data" --image_dims 256 --circle_start 9 --circle_end 10 --add_tri True --add_ellipse True --tri_count 40 --ell_count 10 --size_vary True

6. Create a dataset of 500 images of size=256 x 256 with circle+square+triangle objects of radius=10, side=10, size=12 and circle density=10, square density=10, triangle density=15 and change the intensity of objects to 0.5*std. of the background

To change from square to rectangle objects vary the values in the parameter size, a good starting value (-8,12)

poetry run symmsyn --job_type "multi" --n_image 500 --folder_name "circle_square_tri_data" --image_dims 256 --circle_radius 10 --circle_start 9 --circle_end 10 --add_square True --add_tri True --square_size 10 10 --tri_size 12 12 --sq_count 10 --tri_count 15 --intensity_vary True --intensity_ratio 0.5

7. Create only an image and mask size=256 x 256 with circle+square+triangle objects of radius=10,side=10, size=12 and circle density=10, square density=10, triangle density=15

The image and mask are saved as a file "multi.png"

To change from square to rectangle objects vary the values in the parameter size, a good starting value (-8,12)

poetry run symmsyn --job_type "multi" --image_dims 256 --circle_radius 10 --circle_start 9 --circle_end 10 --add_square True --add_tri True --square_size 10 10 --tri_size 12 12 --sq_count 10 --tri_count 15 --save_file False

=================================================================================================================================================================

Some example images are shown here.

The datasets in the order from left to right: circle objects, circle and ellipse, circle and square, circle, ellipse and triangle objects.

An example of circle, ellipse and triangle object dataset with the variation of signal to noise.