Papers by Masashi Yokota
Augmenting Image Question Answering Dataset by Exploiting Image Captions (L18-1)
Copied to clipboard
| Challenge: | Image question answering requires large amounts of human-annotated data to achieve optimal performance. |
| Approach: | They propose a framework to augment training data by generating additional examples from unannotated pairs of an image and captions. |
| Outcome: | The proposed framework augments training data by generating additional examples from unannotated pairs of an image and captions. |