Research map / Detection and segmentation / 3D detection and driving
INT: Towards Infinite-Frames 3D Detection with an Efficient Framework
Jianyun Xu, Zhenwei Miao, Da Zhang, Hongyu Pan, Kaixuan Liu, Peihan Hao, Jun Zhu, Zhengyang Sun, Hongmin Li, Xin Zhan
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 INT memory bank on 10 frames | CenterPoint-Voxel · Waymo val (L2) | mAPH: 65.5 -> 73.2 | Improved |
| use INT 2 frames instead of CenterPoint 2 frames | CenterPoint-Voxel · Waymo val (L2) | latency (ms): 90.9 -> 74.0 | Improved |
| increase INT frames from 10 to 100 (ablation) | INT-Voxel · nuScenes val | NDS: 65.5 -> 67.3 | Improved |
| add foreground pointcloud fusion (ablation) | INT-Voxel · Waymo val (L2) | mAPH: 65.5 -> 72.3 | Improved |
| add feature map fusion alone (ablation) | INT-Voxel · Waymo val (L2) | mAPH: 65.5 -> 69.1 | Improved |
| add prediction map fusion alone (ablation) | INT-Voxel · Waymo val (L2) | mAPH: 65.5 -> 70.0 | Improved |
| add FM fusion on top of PC fusion (ablation) | INT-Voxel · Waymo val (L2) | mAPH: 72.3 -> 73.5 | Improved |
| add prediction map fusion on top of PC+FM (ablation) | INT-Voxel · Waymo val (L2) | mAPH: 73.5 -> 73.5 | No clear effect |
| add sequence point transformation augmentation (ablation) | INT-Voxel · Waymo val (L2) | mAPH: 63.1 -> 72.1 | Improved |
| add sequence GtAug (ablation) | INT-Voxel · Waymo val (L2) | mAPH: 72.1 -> 73.5 | Improved |
| add dynamic training sequence length (ablation) | INT-Pillar 100 frames · nuScenes val | NDS: 59.3 -> 61.8 | Improved |
| use Concat instead of Add image-style fusion (ablation) | INT-Pillar · nuScenes val | NDS: 42.0 -> 45.3 | Improved |
Closest papers by meaning
- M3SOT: Multi-Frame, Multi-Field, Multi-Space 3D Single Object Tracking (AAAI 2024 · similarity 0.83)
- MAD: Memory-Augmented Detection of 3D Objects (CVPR 2025 · similarity 0.83)
- Cross-Modal Distillation for 2D/3D Multi-Object Discovery from 2D Motion (CVPR 2025 · similarity 0.83)
- ADA-Track: End-to-End Multi-Camera 3D Multi-Object Tracking with Alternating Detection and Association (CVPR 2024 · similarity 0.83)
- FASTer: Focal token Acquiring-and-Scaling Transformer for Long-term 3D Objection Detection (CVPR 2025 · similarity 0.82)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.82)
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.
