v2: Add twice as many images to the dataset and use another model for training ;)
Recommended weight: 0.6 - 0.8
Add more images and use another model for training ;)