The task of visual question answering (VQA) is the task of answering an open-ended text question about a given image. VQA was proposed by Antol and its co-authors in 2015 (https://www.cv-foundation.org/openaccess/content_iccv_2015/papers/Antol_VQA_Visual_Question_ICCV_2015_paper.pdf). This task lies at the intersection of computer vision and natural language processing. It requires the understanding of the image and the parsing and understanding of the text question. Due to its multimodality nature and its well-defined quantitative evaluation metric, VQA is considered an important artificial intelligence task. It also has potential practical applications, including helping the visually impaired.
A few examples of the VQA task are illustrated in the following table:
 |
 |
 |
|
Q: How many giraffes can be seen? A: 2 |
Q: Is the bus door open? A: Yes... |
