Research map / Vision-language and multimodal / Visual question answering
Image-Question-Answer Synergistic Network for Visual Dialog
Check 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 the image-question-answer synergistic re-ranking stage to the generative primary stage (ablation) | Generative visual dialog model · Visual Dialog v1.0 validation | MRR: 49.01 -> 51.62 | Helped |
| Add the image-question-answer synergistic re-ranking stage to the discriminative primary stage (ablation) | Discriminative visual dialog model · Visual Dialog v1.0 validation | MRR: 62.63 -> 63.54 | Helped |
| Use temperature τ=0.50 instead of τ=1.00 (ablation) | Discriminative primary-stage loss · Visual Dialog v1.0 validation | MRR: 61.92 -> 62.51 | Helped |
| Use temperature τ=0.25 instead of τ=0.50 (ablation) | Discriminative primary-stage loss · Visual Dialog v1.0 validation | MRR: 62.51 -> 62.63 | Helped |
| Increase the number of sampled primary-stage candidates M from 10 to 30 (ablation) | Discriminative synergistic stage · Visual Dialog v1.0 validation | MRR: 62.31 -> 63.54 | Helped |
| Increase M from 30 to 40 (ablation) | Discriminative synergistic stage · Visual Dialog v1.0 validation | MRR: 63.54 -> 63.14 | Hurt |
| Increase the number of re-ranked answers N from 10 to 15 (ablation) | Discriminative synergistic stage · Visual Dialog v1.0 validation | MRR: 63.54 -> 63.16 | Hurt |
| Increase the number of re-ranked answers N from 10 to 30 (ablation) | Generative synergistic stage · Visual Dialog v1.0 validation | MRR: 51.62 -> 53.73 | Helped |
| Ensemble 10 models with different seeds and M values | Discriminative synergistic model · Visual Dialog v1.0 test-standard | NDCG: 57.32 -> 57.88 | Helped |
| Use the proposed discriminative model instead of LF-att | Visual dialog answer ranking · Visual Dialog v1.0 test-standard | NDCG: 49.76 -> 57.32 | Helped |
Closest papers by meaning
- Two Can Play This Game: Visual Dialog With Discriminative Question Generation and Answering (CVPR 2018 · similarity 0.92)
- Visual Dialog (CVPR 2017 · similarity 0.92)
- UTC: A Unified Transformer With Inter-Task Contrastive Learning for Visual Dialog (CVPR 2022 · similarity 0.91)
- Modality-Balanced Models for Visual Dialogue (AAAI 2020 · similarity 0.91)
- Visual Reference Resolution using Attention Memory for Visual Dialog (NeurIPS 2017 · similarity 0.90)
- Making the v in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering (CVPR 2017 · 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.