Refining AttnGAN Using Attention on Attention Network.

AttnGAN finds the semantic relation between text and image using an attention network. However, some of the words in the text description remain unattended. We propose a solution called Refined AttnGAN, which contains enhanced attention using Attention on Attention architecture. We apply the mode-seeking function to the network to improve the model diversity. The proposed Architecture is evaluated on the Caltech Birds and Microsoft Coco Dataset. Experimental results demonstrate that our model works very well compared to some state-of-the-art methods.