Image Captioning with Very Scarce Supervised Data: Adversarial Semi-Supervised Learning Approach | Read Paper on Bytez