Research map / Vision-language and multimodal / Visual question answering
Combining Multiple Cues for Visual Madlibs Question Answering
Tatiana Tommasi, Arun Mallya, Bryan A. Plummer, Svetlana Lazebnik, Alexander C. Berg, Tamara L. Berg
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 Places scene features to ImageNet baseline (ablation) | ImageNet fc7 · Visual Madlibs Easy, Scene | accuracy: 87.73 -> 89.04 | Improved |
| add Places scene features to ImageNet baseline (ablation) | ImageNet fc7 · Visual Madlibs Hard, Scene | accuracy: 70.94 -> 73.22 | Improved |
| add Attribute person-box features to baseline (ablation) | ImageNet fc7 · Visual Madlibs Easy, Person's Attribute | accuracy: 53.32 -> 64.50 | Improved |
| add HICO person-box features to baseline (ablation) | ImageNet fc7 · Visual Madlibs Easy, Person's Activity | accuracy: 83.89 -> 87.46 | Improved |
| add Color object features to baseline (ablation) | ImageNet fc7 · Visual Madlibs Easy, Object's Attribute | accuracy: 50.15 -> 59.50 | Improved |
| add object-box ImageNet features to baseline (ablation) | ImageNet fc7 · Visual Madlibs Easy, Object's Affordance | accuracy: 80.56 -> 87.20 | Improved |
| add Color features to object-box ImageNet baseline (ablation) | ImageNet fc7 · Visual Madlibs Easy, Object's Affordance | accuracy: 87.20 -> 82.32 | Hurt |
| add Attribute labels to HICO+MPII label combination (ablation) | HICO+MPII labels · Visual Madlibs Easy, Person's Attribute | accuracy: 59.37 -> 68.43 | Improved |
| use HICO+MPII labels instead of best fc7 cue (ablation) | best single fc7 cue · Visual Madlibs Easy, Interesting/Past/Future/Location | accuracy: small drop | Hurt |
| add Person and Object selection scores to HICO+MPII+Attr CCA (ablation) | HICO+MPII+Attr CCA · Visual Madlibs Hard, Past | accuracy: 58.17 -> 58.29 -> 59.60 | Improved |
| combine all CCA cue scores into learned ensemble (L1) (ablation) | best single cue · Visual Madlibs, average Easy | accuracy: 77.30 -> 79.19 | Improved |
| combine all CCA cue scores into learned ensemble (L1) (ablation) | best single cue · Visual Madlibs, average Hard | accuracy: 60.03 -> 61.52 | Improved |
Closest papers by meaning
- Multi-Level Attention Networks for Visual Question Answering (CVPR 2017 · similarity 0.89)
- An Analysis of Visual Question Answering Algorithms (ICCV 2017 · similarity 0.88)
- Learning Answer Embeddings for Visual Question Answering (CVPR 2018 · similarity 0.88)
- Improved Fusion of Visual and Language Representations by Dense Symmetric Co-Attention for Visual Question Answering (CVPR 2018 · similarity 0.88)
- Multi-Modal Factorized Bilinear Pooling With Co-Attention Learning for Visual Question Answering (ICCV 2017 · similarity 0.88)
- Multi-Question Learning for Visual Question Answering (AAAI 2020 · similarity 0.87)
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.
