Research map / Vision-language and multimodal / Visual question answering
FVQA: Fact-Based Visual Question Answering
Peng Wang, Qi Wu, Chunhua Shen, Anthony Dick, Anton van den Hengel
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 changed | Where | Measured | Effect |
|---|---|---|---|
| use top-3 instead of top-1 predicted question-query mappings (ablation) | FVQA reasoning model · FVQA overall Top-1 | accuracy %: 52.56 -> 56.91 | Improved |
| use ground-truth question-query mapping instead of predicted top-3 (ablation) | FVQA reasoning model · FVQA overall Top-1 | accuracy %: 56.91 -> 63.63 | Improved |
| use explicit KB reasoning instead of LSTM-Question+Image | LSTM baseline · FVQA overall Top-1 | accuracy %: 22.97 -> 56.91 | Improved |
| ensemble LSTM model with knowledge-based model | FVQA reasoning model · FVQA overall Top-1 | accuracy %: 56.91 -> 58.76 | Improved |
| add image input to question-only LSTM | LSTM-Question · FVQA overall Top-1 | accuracy %: 10.45 -> 22.97 | Improved |
| add Pre-VQA extra training data to hierarchical model | Hie-Question+Image · FVQA overall Top-1 | accuracy %: 33.70 -> 43.14 | Improved |
| add Pre-VQA extra training data to LSTM model | LSTM-Question+Image · FVQA overall Top-1 | accuracy %: 22.97 -> 24.98 | Improved |
| add question to SVM image features | SVM-Image · FVQA overall Top-1 | accuracy %: 18.41 -> 18.89 | No clear effect |
| use top-3 mapping on supporting-fact prediction (ablation) | FVQA reasoning model · FVQA supporting-fact Top-1 | accuracy %: 38.76 -> 41.12 | Improved |
| compare WebChild questions with other KB sources for mapping | QQmaping · FVQA | Top-1 mapping accuracy %: 83.15 -> 64.94 | Improved |
| proposed model vs Hie+Pre-VQA on WebChild questions | FVQA reasoning model · FVQA WebChild Top-1 | accuracy %: 48.74 -> 52.85 | Hurt |
Closest papers by meaning
- Answer Them All! Toward Universal Visual Question Answering Models (CVPR 2019 · similarity 0.95)
- OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge (CVPR 2019 · similarity 0.94)
- A-OKVQA: A Benchmark for Visual Question Answering Using World Knowledge (ECCV 2022 · similarity 0.94)
- VQA: Visual Question Answering (IJCV 2016 · similarity 0.94)
- VTQA: Visual Text Question Answering via Entity Alignment and Cross-Media Reasoning (CVPR 2024 · similarity 0.92)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.92)
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.
