Triple Sequence Generative Adversarial Nets for Unsupervised Image Captioning | IEEE Conference Publication | IEEE Xplore