arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2007.04422cs.CVcs.CL
收藏

IQ-VQA:智能可视问答

IQ-VQA: Intelligent Visual Question Answering

Indian Institute of Technology Guwahati(印度理工学院古瓦哈提分校)

Vatsal Goel, Mohit Chandak, Ashish Anand, Prithwijit Guha

摘要:

尽管在可视化问答领域已经取得了巨大的进步,但今天的模型仍然倾向于不一致和脆弱。为此,我们提出了一个模型独立的循环框架,增加了一致性和鲁棒性的任何VQA架构。我们训练我们的模型来回答原始问题,基于答案生成一个含义,然后还学习正确地回答生成的含义。作为循环框架的一部分,我们提出了一种新的隐含生成器,它可以从任何问答对中生成隐含问题。作为未来一致性工作的基线,我们提供了一个新的人类注释VQA-影响数据集。该数据集由约30 k个问题组成,包含3种类型的含义-逻辑等价、必要条件和互斥-由VQA v2.0验证数据集生成。我们表明,我们的框架在基于规则的数据集上将VQA模型的一致性提高了约15%,在VQA-Implications数据集上提高了约7%,鲁棒性提高了约2%,而不会降低其性能。此外,我们还定量地显示了注意力地图的改善,这突出了对视觉和语言的更好的多模态理解。

英文摘要:

Even though there has been tremendous progress in the field of Visual Question Answering, models today still tend to be inconsistent and brittle. To this end, we propose a model-independent cyclic framework which increases consistency and robustness of any VQA architecture. We train our models to answer the original question, generate an implication based on the answer and then also learn to answer the generated implication correctly. As a part of the cyclic framework, we propose a novel implication generator which can generate implied questions from any question-answer pair. As a baseline for future works on consistency, we provide a new human annotated VQA-Implications dataset. The dataset consists of ~30k questions containing implications of 3 types - Logical Equivalence, Necessary Condition and Mutual Exclusion - made from the VQA v2.0 validation dataset. We show that our framework improves consistency of VQA models by ~15% on the rule-based dataset, ~7% on VQA-Implications dataset and robustness by ~2%, without degrading their performance. In addition, we also quantitatively show improvement in attention maps which highlights better multi-modal understanding of vision and language.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑