Visual Question Answering

These are a series of projects we are working on with a broader goal of building AI complete systems. Visual Question Answering is a task that requires combining insights from visual and NLP domains in a deceptively non trivial fashion.

Slack Group
Google Sheet to track progress
Overleaf Doc for update

There are three ideas we are working on :

Improving parsing

Employing additional information such as captions to improve the models

Encouraging agents to learn by asking questions

Parsing

Captioning

Reinforcement Learning