Research map / Video / Video segmentation and tracking
Multimodal Motion Prediction With Stacked Transformers
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 RTS and increase proposals from 6 to 36 | mmTransformer · Argoverse test | MR (%): 15.42 -> 13.00 | Helped |
| Add RTS to a 36-proposal model (ablation) | mmTransformer · Argoverse validation | MR (%): 17.6 -> 9.2 | Helped |
| Add map aggregator to the motion extractor (ablation) | mmTransformer · Argoverse validation | MR (%): 23.3 -> 14.4 | Helped |
| Add social constructor after motion and map modules (ablation) | mmTransformer · Argoverse validation | MR (%): 14.4 -> 10.6 | Helped |
| Use 36 proposals without RTS instead of 6 proposals (ablation) | mmTransformer · Argoverse validation | MR (%): 10.6 -> 17.6 | Hurt |
| Use 36 proposals with RTS instead of 6 proposals without RTS (ablation) | mmTransformer · Argoverse validation | minADE: 0.713 -> 0.721 | No clear effect |
| Use 36 proposals with RTS instead of 6 proposals without RTS (ablation) | mmTransformer · Argoverse validation | MR (%): 10.6 -> 9.2 | Helped |
| Replace K-means partitioning with manual partitioning (ablation) | RTS · Argoverse validation | MR (%): 9.21 -> 9.13 | Helped |
| Replace K-means partitioning with manual partitioning (ablation) | RTS · Argoverse validation | minADE: 0.72 -> 0.73 | Hurt |
| Set M=6 and N=6 rather than M=3 and N=6 (ablation) | RTS · Argoverse validation | MR (%): 11.65 -> 9.21 | Helped |
| Increase regions from M=6 to M=9 with N=6 (ablation) | RTS · Argoverse validation | MR (%): 9.21 -> 9.37 | Hurt |
| Increase proposals per region from N=6 to N=8 with M=6 (ablation) | RTS · Argoverse validation | MR (%): 9.21 -> 9.23 | No clear effect |
Closest papers by meaning
- EDA: Evolving and Distinct Anchors for Multimodal Motion Prediction (AAAI 2024 · similarity 0.86)
- SmartPretrain: Model-Agnostic and Dataset-Agnostic Representation Learning for Motion Prediction (ICLR 2025 · similarity 0.85)
- ModeSeq: Taming Sparse Multimodal Motion Prediction with Sequential Mode Modeling (CVPR 2025 · similarity 0.84)
- MoST: Multi-Modality Scene Tokenization for Motion Prediction (CVPR 2024 · similarity 0.84)
- Motion Diversification Networks (CVPR 2024 · similarity 0.84)
- SEPT: Towards Efficient Scene Representation Learning for Motion Prediction (ICLR 2024 · similarity 0.84)
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.