In this version I used better resolution images and cleaned it up, since it was my first model that I uploaded, I didn't know many things that can optimize the model, so in terms of flexibility it is better, with 52 images with which I trained it.