English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recently, studies of visual question answering have explored variousarchitectures of end-to-end networks and achieved promising results on bothnatural and synthetic datasets, which require explicitly compositionalreasoning. However, it has been argued that these black-box approaches lackinterpretability of results, and thus cannot perform well on generalizationtasks due to overfitting the dataset bias. In this work, we aim to combine thebenefits of both sides and overcome their limitations to achieve an end-to-endinterpretable structural reasoning for general images without the requirementof layout annotations. Inspired by the property of a capsule network that cancarve a tree structure inside a regular convolutional neural network (CNN), wepropose a hierarchical compositional reasoning model called the "Linguisticallydriven Graph Capsule Network", where the compositional process is guided by thelinguistic parse tree. Specifically, we bind each capsule in the lowest layerto bridge the linguistic embedding of a single word in the original questionwith visual evidence and then route them to the same capsule if they aresiblings in the parse tree. This compositional process is achieved byperforming inference on a linguistically driven conditional random field (CRF)and is performed across multiple graph capsule layers, which results in acompositional reasoning process inside a CNN. Experiments on the CLEVR dataset,CLEVR compositional generation test, and FigureQA dataset demonstrate theeffectiveness and composition generalization ability of our end-to-end model.

Linguistically Driven Graph Capsule Network for Visual Question Reasoning