RUBi: Reducing Unimodal Biases for Visual Question Answering - Sorbonne Université Access content directly
Conference Papers Year : 2019

RUBi: Reducing Unimodal Biases for Visual Question Answering

Abstract

Visual Question Answering (VQA) is the task of answering questions about an image. Some VQA models often exploit unimodal biases to provide the correct answer without using the image information. As a result, they suffer from a huge drop in performance when evaluated on data outside their training set distribution. This critical issue makes them unsuitable for real-world settings. We propose RUBi, a new learning strategy to reduce biases in any VQA model. It reduces the importance of the most biased examples, i.e. examples that can be correctly classified without looking at the image. It implicitly forces the VQA model to use the two input modalities instead of relying on statistical regularities between the question and the answer. We leverage a question-only model that captures the language biases by identifying when these unwanted regularities are used. It prevents the base VQA model from learning them by influencing its predictions. This leads to dynamically adjusting the loss in order to compensate for biases. We validate our contributions by surpassing the current state-of-the-art results on VQA-CP v2. This dataset is specifically designed to assess the robustness of VQA models when exposed to different question biases at test time than what was seen during training. Our code is available: github.com/cdancette/rubi.bootstrap.pytorch
Fichier principal
Vignette du fichier
8371-rubi-reducing-unimodal-biases-for-visual-question-answering.pdf (6.16 Mo) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02507524 , version 1 (13-03-2020)

Identifiers

  • HAL Id : hal-02507524 , version 1

Cite

Remi Cadene, Corentin Dancette, Hedi Ben-Younes, Matthieu Cord, Devi Parikh. RUBi: Reducing Unimodal Biases for Visual Question Answering. Neural Information Processing Systems, Dec 2019, Vancouver, Canada. pp.841-852. ⟨hal-02507524⟩
86 View
89 Download

Share

Gmail Facebook X LinkedIn More