Curriculum Learning Effectively Improves Low Data VQA

ALTA 2021 · Narjes Askarian, Ehsan Abbasnejad, Ingrid Zukerman, Wray Buntine, Gholamreza Haffari ·

Visual question answering (VQA) models, in particular modular ones, are commonly trained on large-scale datasets to achieve state-of-the-art performance. However, such datasets are sometimes not available. Further, it has been shown that training these models on small datasets significantly reduces their accuracy. In this paper, we propose curriculum-based learning (CL) regime to increase the accuracy of VQA models trained on small datasets. Specifically, we offer three criteria to rank the samples in these datasets and propose a training strategy for each criterion. Our results show that, for small datasets, our CL approach yields more accurate results than those obtained when training with no curriculum.

PDF Abstract