AI Paper Atlas

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

Learning Visual Question Answering by Bootstrapping Hard Attention

Mateusz Malinowski, Carl Doersch, Adam Santoro, Peter Battaglia

ECCV 2018 · Vision-language and multimodal › Visual question answering · approach: Attention and prior-robust VQA · 102 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
select only 16% of cells by L2 norm instead of all cells (sum pooling) (ablation)sum pooling · VQA-CP v2overall accuracy: 27.96 -> 26.99No clear effect
select 48% of cells with hard attention instead of all cells (pairwise aggregation) (ablation)non-local pairwise · VQA-CP v2overall accuracy: 28.06 -> 28.23Improved
use soft attention instead of sum poolingsum pooling · VQA-CP v2overall accuracy: 27.96 -> 26.60No clear effect
add adaptive threshold selection (AdaHAN) to pairwise aggregation instead of fixed 48% cellsHAN+pairwise · VQA-CP v2overall accuracy: 28.23 -> 28.65Improved
add adaptive selection (AdaHAN) to sum pooling vs HAN 32%HAN+sum · VQA-CP v2overall accuracy: 27.43 -> 27.40No clear effect
add the two alignment layers before attention (HAN+sum 25%) (ablation)HAN+sum · VQA-CP v2overall accuracy: 26.38 -> 26.99Improved
add the two alignment layers before attention (HAN+sum 75%) (ablation)HAN+sum · VQA-CP v2overall accuracy: 26.82 -> 27.94Improved
replace soft attention with HAN when training CNN from scratchsimple-SAN · VQA-CP v2accuracy: 21% -> 24%Improved
use HAN+sum instead of soft attention SANSAN · CLEVR testoverall accuracy: 68.5 -> 89.7Improved
add relation network aggregation to HANHAN+sum · CLEVR testoverall accuracy: 89.7 -> 93.9Improved
scale to larger model and higher resolution (HAN+RN++)HAN+RN+ · CLEVR testoverall accuracy: 96.9 -> 98.8Improved
use hard attention (25% cells) with relation network instead of full RNRelation Network · CLEVR validation, 12 h trainingvalidation accuracy: 70% -> 90%Improved

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