Abstract: Vision-to-language tasks require a unified semantic understanding of visual content. However, the information contained in image/video is essentially ambiguous on two perspectives manifested ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results