I attached the image for 1st step and boxing is very strange ? Any idea ? Also would you consider to upgrade your codes according to 2017 Deep Learning techniques ? :) <img width="758" alt="screen shot 2017-09-29 at 21 57 08" src="https://user-images.githubusercontent.com/17505439/31031477-552e52e4-a561-11e7-91ed-526dd72542c2.png">