This image dataset generator was made for a project in DAE Research, where I worked on it as a sole programmer.
When working on it, the wanted features were still undecided, so I focused on developing this application to be flexible and easy to expand. The system works through rules that have been programmed, these rules go from what and how many objects get spawned, to things like how much noise gets added to the final image.

To expand the application new rules can be added, and can be assigned to different moments in the image generation procedure (like before an image gets generated, rules specific to certain spawned objects, image postprocessing rules,....) Then the UI for each rule can also be created with just one to a few lines of code.

While the application was developed with the idea that rules get added when needed, multiple rules were already added. These include:
-How many images get created
-Resolution of images
-Using a random or a specific seed
-Range of amount of objects that get spawned
-Range of position and rotation of objects
-Masking of objects
-Save as png or jpg + jpg quality range
-Box annotation
-...