Research map / Vision-language and multimodal / Human and embodied modeling
Dense Captioning With Joint Inference and Visual Context
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 |
|---|---|---|---|
| Use two separate LSTMs for localization and captioning instead of a shared caption LSTM (ablation) | Dense captioning model · Visual Genome V1.0 | mAP: 6.47 -> 8.03 | Helped |
| Use two separate LSTMs for localization and captioning rather than the baseline architecture (ablation) | Dense captioning model · Visual Genome V1.0 | mAP: 6.85 -> 8.03 | Helped |
| Use a shared caption LSTM to predict the box offset (ablation) | Dense captioning model · Visual Genome V1.0 | mAP: 6.85 -> 6.47 | Hurt |
| Add region features to the shared LSTM output for box prediction (ablation) | Dense captioning model · Visual Genome V1.0 | mAP: 6.85 -> 6.83 | No clear effect |
| Add context using late-fusion multiplication to T-LSTM (ablation) | T-LSTM · Visual Genome V1.0 | mAP: 8.03 -> 8.60 | Helped |
| Use late-fusion rather than early-fusion for context (ablation) | Integrated dense captioning models · Visual Genome V1.0 | mAP: T-LSTM: 8.24 -> 8.49 (concatenation); 8.19 -> 8.60 (multiplication); 8.16 -> 8.53 (summation) | Helped |
| Add late-fusion context to T-LSTM using concatenation (ablation) | T-LSTM · Visual Genome V1.2 | mAP: 8.16 -> 9.03 | Helped |
| Use 300 rather than 100 proposals (ablation) | T-LSTM-mult · Visual Genome V1.0 | mAP: 8.67 -> 9.31 | Helped |
| Use 300 rather than 100 proposals (ablation) | T-LSTM-concat · Visual Genome V1.2 | mAP: 9.47 -> 9.96 | Helped |
| Use the proposed model rather than Johnson et al. | Dense captioning · Visual Genome V1.0 | mAP: 5.39 -> 9.31 | Helped |
Closest papers by meaning
- A Picture is Worth More Than 77 Text Tokens: Evaluating CLIP-Style Models on Dense Captions (CVPR 2024 · similarity 0.91)
- Dense Video Object Captioning from Disjoint Supervision (ICLR 2025 · similarity 0.90)
- Benchmarking Large Vision-Language Models via Directed Scene Graph for Comprehensive Image Captioning (CVPR 2025 · similarity 0.89)
- DenseFusion-1M: Merging Vision Experts for Comprehensive Multimodal Perception (NeurIPS 2024 · similarity 0.88)
- FlexCap: Describe Anything in Images in Controllable Detail (NeurIPS 2024 · similarity 0.88)
- VisMin: Visual Minimal-Change Understanding (NeurIPS 2024 · similarity 0.88)
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.