The LoRa was trained on a large number of similar images, 150 images +/-. each image has a water element, a tree as the focal point, lots of roots and twigs and leaves etc.