Research map / Vision-language and multimodal / Visual question answering
Structured Attentions for Visual Question Answering
Chen Zhu, Yanpeng Zhao, Shuaiyi Huang, Kewei Tu, Yi Ma
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 MF structured attention to sigmoid attention (ablation) | SIG · CLEVR validation (res5c) | overall accuracy: 70.52 -> 73.14 | Improved |
| add LBP structured attention to sigmoid attention (ablation) | SIG · CLEVR validation (res5c) | overall accuracy: 70.52 -> 72.30 | Improved |
| add MF structured attention (ablation) | SIG · CLEVR validation (res4b22) | overall accuracy: 75.32 -> 76.65 | Improved |
| combine LBP with SIG attention (ablation) | LBP · CLEVR validation (res4b22) | overall accuracy: 76.21 -> 77.97 | Improved |
| combine MF with SIG attention (ablation) | MF · CLEVR validation (res5c) | overall accuracy: 73.14 -> 73.19 | No clear effect |
| use MF-SIG instead of SAN | VQA model · CLEVR test | overall accuracy: 68.5 -> 77.57 | Improved |
| use res4b22 instead of res5c features | SIG · CLEVR validation | overall accuracy: 70.52 -> 75.32 | Improved |
| increase inference steps T from 1 to 3 (ablation) | MF-SIG · VQA test-dev | overall accuracy: 65.90 -> 66.00 | No clear effect |
| increase inference steps T from 1 to 4 (ablation) | MF-SIG · VQA test-dev | overall accuracy: 65.90 -> 65.81 | No clear effect |
| increase inference steps T from 1 to 3 (ablation) | LBP-SIG · VQA test-dev | overall accuracy: 65.93 -> 65.81 | Hurt |
| add Visual Genome external data | MF-SIG · VQA test-dev | overall accuracy: 66.00 -> 67.19 | Improved |
| use MF-SIG instead of MLB under VG data | VQA model · VQA test-dev | overall accuracy: 65.84 -> 67.19 | Improved |
Closest papers by meaning
- Multi-Level Attention Networks for Visual Question Answering (CVPR 2017 · similarity 0.93)
- Hierarchical Question-Image Co-Attention for Visual Question Answering (NeurIPS 2016 · similarity 0.93)
- Graph-Structured Representations for Visual Question Answering (CVPR 2017 · similarity 0.93)
- An Analysis of Visual Question Answering Algorithms (ICCV 2017 · similarity 0.92)
- Re-Attention for Visual Question Answering (AAAI 2020 · similarity 0.92)
- Relation-Aware Graph Attention Network for Visual Question Answering (ICCV 2019 · 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.
