This dataset consists of 247 360x202 colour images. 100 images are for training, 72 images are for validation and 75 for test.