Don't Just Assume; Look and Answer: Overcoming Priors for Visual Question Answering

@article{Agrawal2018DontJA,
  title={Don't Just Assume; Look and Answer: Overcoming Priors for Visual Question Answering},
  author={Aishwarya Agrawal and Dhruv Batra and Devi Parikh and Aniruddha Kembhavi},
  journal={2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  year={2018},
  pages={4971-4980}
}
  • Aishwarya Agrawal, Dhruv Batra, +1 author Aniruddha Kembhavi
  • Published 2018
  • Computer Science
  • 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition
  • A number of studies have found that today's Visual Question Answering (VQA) models are heavily driven by superficial correlations in the training data and lack sufficient image grounding. [...] Key Result Second, we propose a novel Grounded Visual Question Answering model (GVQA) that contains inductive biases and restrictions in the architecture specifically designed to prevent the model from 'cheating' by primarily relying on priors in the training data. Specifically, GVQA explicitly disentangles the…Expand Abstract

    Citations

    Publications citing this paper.
    SHOWING 1-10 OF 126 CITATIONS

    Inverse Visual Question Answering: A New Benchmark and VQA Diagnosis Tool

    VIEW 1 EXCERPT
    CITES BACKGROUND

    Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering

    VIEW 5 EXCERPTS
    CITES BACKGROUND & METHODS

    Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering

    VIEW 5 EXCERPTS
    CITES BACKGROUND & METHODS

    Counterfactual Samples Synthesizing for Robust Visual Question Answering

    VIEW 7 EXCERPTS
    CITES BACKGROUND & METHODS
    HIGHLY INFLUENCED

    Answer Them All! Toward Universal Visual Question Answering Models

    VIEW 16 EXCERPTS
    CITES BACKGROUND & METHODS
    HIGHLY INFLUENCED

    P $\approx$ NP, at least in Visual Question Answering

    VIEW 4 EXCERPTS
    CITES BACKGROUND & METHODS
    HIGHLY INFLUENCED

    Adversarial Regularization for Visual Question Answering: Strengths, Shortcomings, and Side Effects

    VIEW 7 EXCERPTS
    CITES METHODS & BACKGROUND
    HIGHLY INFLUENCED

    FILTER CITATIONS BY YEAR

    2017
    2020

    CITATION STATISTICS

    • 37 Highly Influenced Citations

    • Averaged 41 Citations per year from 2018 through 2020

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 45 REFERENCES

    Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering

    VIEW 7 EXCERPTS

    An Analysis of Visual Question Answering Algorithms

    VIEW 1 EXCERPT

    An Empirical Evaluation of Visual Question Answering for Novel Objects

    VIEW 1 EXCERPT

    Cross-Dataset Adaptation for Visual Question Answering

    VIEW 1 EXCERPT

    Visual7W: Grounded Question Answering in Images

    VIEW 1 EXCERPT

    Answer-Type Prediction for Visual Question Answering

    VIEW 1 EXCERPT