Research map / Vision-language and multimodal / Visual question answering
Visual Question Answering With Memory-Augmented Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 external memory to the ResNet model trained on the top 3000 answers (ablation) | ResNet VQA model · VQA test-dev, multiple-choice | All accuracy: 68.6 -> 69.5 | Helped |
| Add external memory to the ResNet model trained on the top 3000 answers (ablation) | ResNet VQA model · VQA test-dev, open-ended | All accuracy: 62.5 -> 63.8 | Helped |
| Add external memory to the VGGNet model trained on the top 1000 answers (ablation) | VGGNet VQA model · VQA test-dev, multiple-choice | All accuracy: 65.4 -> 66.0 | Helped |
| Add external memory to the VGGNet model trained on the top 1000 answers (ablation) | VGGNet VQA model · VQA test-dev, open-ended | All accuracy: 60.4 -> 61.3 | Helped |
| Add backward LSTM to the forward LSTM question encoder (ablation) | CNN+LSTM baseline · VQA test-dev, multiple-choice | All accuracy: 63.6 -> 64.3 | Helped |
| Add co-attention over image regions and question words (ablation) | Bidirectional LSTM model · VQA test-dev, multiple-choice | All accuracy: 64.3 -> 65.9 | Helped |
| Add external memory (ablation) | Co-attention model · VQA test-dev, multiple-choice | All accuracy: 65.9 -> 67.8 | Helped |
| Increase the answer vocabulary from top 1000 to top 3000 without external memory (ablation) | ResNet VQA model · VQA test-dev, multiple-choice | All accuracy: 68.4 -> 68.6 | Mixed |
| Train the Visual 7W model with additional VQA questions | Visual 7W model · Visual 7W Telling | Overall accuracy: 59.4 -> 62.8 | Helped |
Closest papers by meaning
- Dynamic Memory Networks for Visual and Textual Question Answering (ICML 2016 · similarity 0.91)
- Learning Visual Knowledge Memory Networks for Visual Question Answering (CVPR 2018 · similarity 0.91)
- Multi-Level Attention Networks for Visual Question Answering (CVPR 2017 · similarity 0.91)
- Tips and Tricks for Visual Question Answering: Learnings From the 2017 Challenge (CVPR 2018 · similarity 0.90)
- Ask Me Anything: Free-Form Visual Question Answering Based on Knowledge From External Sources (CVPR 2016 · similarity 0.90)
- Learning Answer Embeddings for Visual Question Answering (CVPR 2018 · similarity 0.90)
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.