Experimenting with Self-Supervision using Rotation Prediction for Image Captioning