This version has the same size of dataset, but it changes it's learning attributes, doing 20 steps instead of 10
RakkunVT model!
Trained with just 27 images, but works very well
