Research map / Video / Video-language understanding
Motion Guided Region Message Passing for Video Captioning
Shaoxiang Chen, Yu-Gang Jiang
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 Recurrent Region Attention without message passing (ablation) | baseline LSTM · MSR-VTT | CIDEr: 42.3 -> 43.1 | No clear effect |
| add motion guided cross-frame message passing (MGCMP) (ablation) | RRA model · MSR-VTT | CIDEr: 43.1 -> 49.9 | Improved |
| add ATGD temporal relation modeling (ablation) | RRA+MGCMP model · MSR-VTT | CIDEr: 49.9 -> 51.4 | Improved |
| add motion guidance to MGCMP (ablation) | MGCMP · MSR-VTT | CIDEr: 46.7 -> 51.4 | Improved |
| add cross-frame message propagation (ablation) | MGCMP · MSR-VTT | CIDEr: 50.6 -> 51.4 | Improved |
| add region diversity loss Ldiv (ablation) | RRA · MSR-VTT | CIDEr: 49.7 -> 51.4 | Improved |
| replace object detector features with RRA region features (ablation) | MGCMP · MSR-VTT | CIDEr: 51.3 -> 51.4 | No clear effect |
| replace object detector with RRA in HVTG | HVTG · Charades-STA | Recall IoU=0.5: 47.27 -> 42.31 | Hurt |
| use 3D ResNeXt-101 instead of C3D motion features | MGCMP · MSVD | CIDEr: 90.7 -> 98.5 | Improved |
| use 3D ResNeXt-101 instead of C3D motion features | MGCMP · MSR-VTT | CIDEr: 50.1 -> 51.4 | Improved |
| region attention instead of object detection for inference speed | MGCMP vs ORG-TRL · 12 s video | FPS: 12.01 -> 65.22 | Improved |
| use lambda 0.1 for attention softness | RRA · MSR-VTT | CIDEr: lambda=0.1 achieves a nice balance | Improved |
Closest papers by meaning
- Motion Guided Spatial Attention for Video Captioning (AAAI 2019 · similarity 0.91)
- Dense Video Object Captioning from Disjoint Supervision (ICLR 2025 · similarity 0.89)
- Fine-Grained Captioning of Long Videos through Scene Graph Consolidation (ICML 2025 · similarity 0.88)
- Event-Equalized Dense Video Captioning (CVPR 2025 · similarity 0.86)
- Do You Remember? Dense Video Captioning with Cross-Modal Memory Retrieval (CVPR 2024 · similarity 0.86)
- Streaming Dense Video Captioning (CVPR 2024 · similarity 0.86)
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.
