AI Paper Atlas

Research map / Vision-language and multimodal / Visual question answering

Straight to the Facts: Learning Knowledge Base Retrieval for Factual Visual Question Answering

Medhini Narasimhan, Alexander G. Schwing

ECCV 2018 · Vision-language and multimodal › Visual question answering · approach: Attention and prior-robust VQA · 107 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add hard negative mining (three iterations) to fact scoring training (ablation)Ours - Question + Visual Concepts · FVQA supporting fact retrievaltop-1 precision: 20.17 -> 64.5Improved
add predicted visual concepts as input to the fact-scoring MLP (ablation)Ours · FVQA answer accuracytop-1 accuracy: 26.68 -> 60.30Improved
add predicted visual concepts as input to the fact-scoring MLP (ablation)Ours · FVQA supporting fact retrievaltop-1 precision: 28.98 -> 62.30Improved
drop image features, keep question and visual concepts only (ablation)Ours · FVQA answer accuracytop-1 accuracy: 60.30 -> 62.20Improved
drop image features, keep question and visual concepts only (ablation)Ours · FVQA supporting fact retrievaltop-1 precision: 62.30 -> 64.50Improved
learned embedding-based relation prediction instead of FVQA baselinerelation prediction · FVQA relation predictiontop-1 accuracy: 64.94 -> 75.4Improved
learned embedding retrieval instead of FVQA ensemble for answersanswer prediction · FVQA answer accuracytop-1 accuracy: 58.76 -> 62.20Improved
learned embedding retrieval instead of FVQA for supporting factsfact retrieval · FVQA supporting fact retrievaltop-1 precision: 41.12 -> 64.50Improved
use ground-truth information instead of predictedOurs - Question + Visual Concepts · FVQA answer accuracytop-1 accuracy: 62.20 -> 70.34Improved
second hard negative mining iteration (84,563 hard negatives) (ablation)Ours · FVQA supporting fact retrievaltop-1 precision: 20.17 -> 38.65Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy