All the images from the 34 classes were considered upon calculation of the 80-20 proportion but only 14 classes with 343 images (20%) were preferred as testing dataset obtained from the new set of stool samples.