Research map / Vision-language and multimodal / Visual question answering
Scene Text Visual Question Answering
Ali Furkan Biten, Ruben Tito, Andres Mafla, Lluis Gomez, Marcal Rusinol, Ernest Valveny, C.V. Jawahar, Dimosthenis Karatzas
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 |
|---|---|---|---|
| add PHOC scene-text features from a retrieval model | SAN(LSTM) 5k cls · ST-VQA Task 1 | accuracy (%): 7.78 -> 10.34 | Improved |
| add PHOC scene-text features | SAAA 5k cls · ST-VQA Task 1 | accuracy (%): 6.66 -> 7.41 | Improved |
| add PHOC scene-text features | SAAA 1k cls · ST-VQA Task 1 | accuracy (%): 6.36 -> 6.66 | Improved |
| add PHOC scene-text features | SAAA 19k cls · ST-VQA Task 1 | accuracy (%): 6.13 -> 6.36 | Improved |
| increase answer vocabulary from 1k to 5k classes | SAAA · ST-VQA Task 1 | accuracy (%): 6.36 -> 6.66 | Improved |
| increase answer vocabulary from 5k to 19k classes | SAAA · ST-VQA Task 1 | accuracy (%): 6.66 -> 6.13 | Hurt |
| use CNN instead of LSTM question encoder | SAN+STR 5k cls · ST-VQA Task 1 | accuracy (%): 10.34 -> 10.46 | No clear effect |
| add visual features to question and STR features | QA+STR vs SAAA+STR 19k cls · ST-VQA Task 1 | accuracy (%): 4.65 -> 6.36 | Improved |
| use VQA model with STR instead of question-blind scene text retrieval | SAN(LSTM)+STR vs STR (retrieval) · ST-VQA Task 1 | accuracy (%): 13.78 -> 10.34 | Hurt |
Closest papers by meaning
- Locate Then Generate: Bridging Vision and Language with Bounding Box for Scene-Text VQA (AAAI 2023 · similarity 0.93)
- On the General Value of Evidence, and Bilingual Scene-Text Visual Question Answering (CVPR 2020 · similarity 0.93)
- Visual Question Answering as Reading Comprehension (CVPR 2019 · similarity 0.92)
- VQA: Visual Question Answering (IJCV 2016 · similarity 0.92)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.91)
- VTQA: Visual Text Question Answering via Entity Alignment and Cross-Media Reasoning (CVPR 2024 · similarity 0.91)
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.
