Research map / 3D vision / 3D generation
MIM4D: Masked Modeling with Multi-View Video for Autonomous Driving Representation Learning
Jialv Zou, Bencheng Liao, Qian Zhang, Wenyu Liu, Xinggang Wang
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 |
|---|---|---|---|
| increase time window from 1 to 5 frames (ablation) | MIM4D · UVTR ResNet50 nuScenes | mAP: 18.2 -> 20.1 | Improved |
| increase time window from 1 to 5 frames (ablation) | MIM4D · UVTR ResNet50 nuScenes | NDS: 22.2 -> 23.5 | Improved |
| add temporal modeling with long-term and short-term transformers (ablation) | MIM4D · UVTR ResNet50 nuScenes | mAP: 18.2 -> 20.1 | Improved |
| add long-term branch to short-term branch (ablation) | MIM4D · UVTR ResNet50 nuScenes | mAP: 18.3 -> 20.1 | Improved |
| add short-term branch to long-term branch (ablation) | MIM4D · UVTR ResNet50 nuScenes | mAP: 19.5 -> 20.1 | Improved |
| use transformer temporal modeling instead of Warp-Cat (ablation) | MIM4D · UVTR ResNet50 nuScenes | mAP: 18.9 -> 20.1 | Improved |
| use short-term transformer only vs none (ablation) | MIM4D · UVTR ResNet50 nuScenes | mAP: 18.2 -> 18.3 | No clear effect |
| pre-train backbone with MIM4D instead of UniPAD | UVTR ConvNeXt-S · nuScenes detection | mAP: 31.1 -> 32.2 | Improved |
| pre-train with MIM4D instead of ImageNet | UVTR ConvNeXt-S · nuScenes detection | mAP: 23.0 -> 32.2 | Improved |
| MIM4D pre-trained backbone | CVT · BEV segmentation Setting 2 | IoU (%): 33.4 -> 36.3 | Improved |
| MIM4D pre-trained backbone | CVT · BEV segmentation Setting 1 | IoU (%): 37.3 -> 39.5 | Improved |
| MIM4D pre-trained backbone | BEVDet4D · nuScenes val detection | mAP: 31.4 -> 32.5 | Improved |
Closest papers by meaning
- M-BEV: Masked BEV Perception for Robust Autonomous Driving (AAAI 2024 · similarity 0.89)
- MaskGWM: A Generalizable Driving World Model with Video Mask Reconstruction (CVPR 2025 · similarity 0.87)
- CorrBEV: Multi-View 3D Object Detection by Correlation Learning with Multi-modal Prototypes (CVPR 2025 · similarity 0.87)
- UniPAD: A Universal Pre-training Paradigm for Autonomous Driving (CVPR 2024 · similarity 0.86)
- DriveDreamer4D: World Models Are Effective Data Machines for 4D Driving Scene Representation (CVPR 2025 · similarity 0.86)
- VisionPAD: A Vision-Centric Pre-training Paradigm for Autonomous Driving (CVPR 2025 · 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.
