Answering Visual What-If Questions: From Actions to Predicted Scene Descriptions
This work addresses a novel problem in scene understanding for AI systems, enabling them to answer visual what-if questions about future states, which is incremental as it extends existing scene description and question answering tasks.
The paper tackles the problem of predicting future scene states conditioned on hypothetical actions, proposing a hybrid model that integrates a physics engine into a question answering architecture to anticipate object-object interactions. It demonstrates first results on this challenging task, outperforming fully data-driven end-to-end learning baselines.
In-depth scene descriptions and question answering tasks have greatly increased the scope of today's definition of scene understanding. While such tasks are in principle open ended, current formulations primarily focus on describing only the current state of the scenes under consideration. In contrast, in this paper, we focus on the future states of the scenes which are also conditioned on actions. We posit this as a question answering task, where an answer has to be given about a future scene state, given observations of the current scene, and a question that includes a hypothetical action. Our solution is a hybrid model which integrates a physics engine into a question answering architecture in order to anticipate future scene states resulting from object-object interactions caused by an action. We demonstrate first results on this challenging new problem and compare to baselines, where we outperform fully data-driven end-to-end learning approaches.