Research map / Vision-language and multimodal / Visual question answering
Progressive Attention Memory Network for Movie Story Question Answering
Junyeong Kim, Minuk Ma, Kyungsu Kim, Sungjin Kim, Chang D. Yoo
Check 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 progressive attention that updates memory using question and answer cues (ablation) | PAMN · MovieQA validation | accuracy: 42.03 -> 43.34 | Improved |
| Add multiple attention hops (ablation) | PAMN · MovieQA validation | accuracy: 42.67 -> 43.34 | Improved |
| Use question-conditioned dynamic modality fusion instead of mean fusion (ablation) | PAMN · MovieQA validation | accuracy: 42.09 -> 43.34 | Improved |
| Use dynamic modality fusion instead of MCB fusion (ablation) | PAMN · MovieQA validation | accuracy: 42.89 -> 43.34 | Improved |
| Use dynamic modality fusion instead of MFB fusion (ablation) | PAMN · MovieQA validation | accuracy: 42.55 -> 43.34 | Improved |
| Use dynamic modality fusion instead of Tucker fusion (ablation) | PAMN · MovieQA validation | accuracy: 42.89 -> 43.34 | Improved |
| Add both Mu and Mg belief corrections (ablation) | PAMN · MovieQA validation | accuracy: 39.50 -> 43.34 | Improved |
| Add Mg belief correction (ablation) | PAMN · MovieQA validation | accuracy: 41.76 -> 43.34 | Improved |
| Add Mu belief correction (ablation) | PAMN · MovieQA validation | accuracy: 40.86 -> 43.34 | Improved |
| Increase attention hops from 1/1 to 2/2 (ablation) | PAMN with pooling 24/16 and correction weights 1/0.5 · MovieQA validation | accuracy: 42.67 -> 43.34 | Improved |
| Increase attention hops from 2/2 to 3/3 (ablation) | PAMN with pooling 24/16 and correction weights 1/0.5 · MovieQA validation | accuracy: 43.34 -> 42.55 | Hurt |
| Use pooling size/stride 80/60 rather than 24/16 (ablation) | PAMN with 1/1 hops and correction weights 1/0.5 · MovieQA validation | accuracy: 42.67 -> 41.20 | Hurt |
Closest papers by meaning
- Movie Question Answering: Remembering the Textual Cues for Layered Visual Contents (AAAI 2018 · similarity 0.91)
- A Read-Write Memory Network for Movie Story Understanding (ICCV 2017 · similarity 0.89)
- Multimodal Dual Attention Memory for Video Story Question Answering (ECCV 2018 · similarity 0.88)
- Structured Two-Stream Attention Network for Video Question Answering (AAAI 2019 · similarity 0.87)
- Knowledge-Based Video Question Answering with Unsupervised Scene Descriptions (ECCV 2020 · similarity 0.87)
- MovieQA: Understanding Stories in Movies Through Question-Answering (CVPR 2016 · 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.
