Research map / Vision-language and multimodal / Visual question answering
Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answering
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 |
|---|---|---|---|
| Replace ResNet spatial features with bottom-up Faster R-CNN salient-region features | Up-Down image captioning model · MSCOCO Karpathy test split | CIDEr: 105.4 -> 113.5 (cross-entropy); 111.1 -> 120.1 (CIDEr optimization) | Helped |
| Replace ResNet spatial features with bottom-up Faster R-CNN salient-region features | Up-Down VQA model · VQA v2.0 validation | Overall accuracy: 59.4 -> 63.2 | Helped |
| Replace ResNet features with bottom-up Faster R-CNN salient-region features | Up-Down captioning model · MSCOCO Karpathy test split | SPICE: Objects: 35.4 -> 37.1 | Helped |
| Replace ResNet features with bottom-up Faster R-CNN salient-region features | Up-Down captioning model · MSCOCO Karpathy test split | SPICE: Attributes: 8.6 -> 9.2 | Helped |
| Replace ResNet features with bottom-up Faster R-CNN salient-region features | Up-Down captioning model · MSCOCO Karpathy test split | SPICE: Relations: 5.3 -> 5.8 | Helped |
| Use 1×1 mean-pooled ResNet features instead of 7×7 spatial features (ablation) | ResNet VQA baseline · VQA v2.0 validation | Overall accuracy: 56.3 vs 59.4 | Hurt |
| Use 14×14 ResNet spatial features instead of 7×7 features (ablation) | ResNet VQA baseline · VQA v2.0 validation | Overall accuracy: 59.4 -> 57.9 | Hurt |
Closest papers by meaning
- Hierarchical Question-Image Co-Attention for Visual Question Answering (NeurIPS 2016 · similarity 0.91)
- Multi-Level Attention Networks for Visual Question Answering (CVPR 2017 · similarity 0.90)
- Dynamic Capsule Attention for Visual Question Answering (AAAI 2019 · similarity 0.90)
- Re-Attention for Visual Question Answering (AAAI 2020 · similarity 0.90)
- FOCUS: Internal MLLM Representations for Efficient Fine-Grained Visual Question Answering (NeurIPS 2025 · similarity 0.89)
- Improved Fusion of Visual and Language Representations by Dense Symmetric Co-Attention for Visual Question Answering (CVPR 2018 · similarity 0.89)
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.