Preprocessing for Faster RCNN

I have found a workaround. Check this. See if it works