Randomly select a background image and object images, and paste them around to create new images. Useful (one would hope) for generating training data for neural networks/deep learning.
python3 imagesim.py -n N [-e E] [-s] [-b <bgs>] [-c <cs>] -o <out>
Where
- N is number of images to generate
- E is the max number of objects per image (default 6)
- s is whether to use only a single class of objects per image, or allow a mixture (default: mixture)
- bgs is a directory of background images (default "backgrounds")
- cs is a directory containing directories of classes (default "crops")
- out is the output directory