Version One - Training data from the following dataset: https://www.kaggle.com/datasets/vishmiperera/children-drawings
CANNY edge detection used on all images in the data folder above 512/512px. Objects with high contour in the image were extracted, scaled to 512/512px and compiled in a new directory. 15 images chosen at random for the training set
(this may be corrupted as a changed the file name, couldn't get it to work in run diffusion)