Research map / Vision-language and multimodal / Visual question answering
Multi-Modal Factorized Bilinear Pooling With Co-Attention Learning for Visual Question Answering
Zhou Yu, Jun Yu, Jianping Fan, Dacheng Tao
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 l2 normalization after MFB output (ablation) | MFB (k=5, o=1000) · VQA test-dev Open-Ended | accuracy: 57.7 -> 60.9 | Improved |
| add power normalization after MFB output (ablation) | MFB (k=5, o=1000) · VQA test-dev Open-Ended | accuracy: 60.4 -> 60.9 | Improved |
| add both power and l2 normalization (ablation) | MFB (k=5, o=1000) · VQA test-dev Open-Ended | accuracy: 57.3 -> 60.9 | Improved |
| increase factor k from 1 to 5 at fixed ko=5000 (ablation) | MFB · VQA test-dev Open-Ended | accuracy: 60.4 -> 60.9 | Improved |
| increase factor k from 5 to 10 (ko=5000) (ablation) | MFB · VQA test-dev Open-Ended | accuracy: 60.9 -> 60.6 | No clear effect |
| increase output dimension o beyond 1000 (2000, 4000) (ablation) | MFB (k=5) · VQA test-dev Open-Ended | accuracy: 60.9 -> 60.7 -> 60.4 | Hurt |
| reduce output dimension o to 200 (ablation) | MFB (k=5) · VQA test-dev Open-Ended | accuracy: 60.9 -> 59.8 | Hurt |
| replace MCB/MLB fusion with MFB (ablation) | bilinear fusion baseline · VQA test-dev Open-Ended | accuracy: 59.8 -> 60.9 | Improved |
| add question attention (co-attention) to image attention (ablation) | MFB+Att · VQA test-dev Open-Ended | accuracy: 64.6 -> 65.1 | Improved |
| add image attention (ablation) | MFB · VQA test-dev Open-Ended | accuracy: 62.2 -> 64.6 | Improved |
| concatenate pre-trained GloVe word vectors (ablation) | MFB+CoAtt · VQA test-dev Open-Ended | accuracy: 65.1 -> 65.9 | Improved |
| add Visual Genome training data (ablation) | MFB+CoAtt+GloVe · VQA test-dev Open-Ended | accuracy: 65.9 -> 66.9 | Improved |
Closest papers by meaning
- Deep Modular Co-Attention Networks for Visual Question Answering (CVPR 2019 · similarity 0.92)
- Deep Attention Neural Tensor Network for Visual Question Answering (ECCV 2018 · similarity 0.91)
- An Analysis of Visual Question Answering Algorithms (ICCV 2017 · similarity 0.91)
- Co-Attending Free-Form Regions and Detections With Multi-Modal Multiplicative Feature Embedding for Visual Question Answering (AAAI 2018 · similarity 0.91)
- Multi-Question Learning for Visual Question Answering (AAAI 2020 · similarity 0.91)
- Improved Fusion of Visual and Language Representations by Dense Symmetric Co-Attention for Visual Question Answering (CVPR 2018 · 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.
