Learning Real-World Robot Policies by Dreaming