Weakly-Supervised Generation and Grounding of Visual Descriptions with Conditional Generative Models | IEEE Conference Publication | IEEE Xplore