Research map / Point clouds / Point cloud classification
Point Primitive Transformer for Long-Term 4D Point Cloud Video Understanding
Hao Wen, Yunze Liu, Jingwei Huang, Bo Duan, Li Yi
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 |
|---|---|---|---|
| use hierarchical PPTr (1 frame) instead of P4Transformer (3 frames) | P4Transformer · Synthia 4D | mIoU: 83.16 -> 83.85 | Improved |
| add 30-frame primitive memory pool instead of 1 frame | PPTr · Synthia 4D | mIoU: 83.85 -> 84.49 | Improved |
| add 30-frame primitive memory pool instead of 3 frames | PPTr · HOI4D | mIoU: 68.54 -> 70.13 | Improved |
| replace P4Transformer with PPTr (1 frame) | P4Transformer · HOI4D | mIoU: 59.61 -> 68.07 | Improved |
| replace P4Transformer with PPTr (24 frames) | P4Transformer · MSR-Action3D | accuracy: 90.94 -> 92.33 | Improved |
| add intra-primitive transformer (ablation) | PPTr · MSR-Action3D | accuracy: 16.73 accuracy drop without it | Improved |
| add inter-primitive (primitive) transformer (ablation) | PPTr · MSR-Action3D | accuracy: 1.39 accuracy drop without it | Improved |
| use primitive planes instead of K-means clusters as mid-level representation (ablation) | PPTr · Synthia 4D | mIoU: 80.70 -> 84.49 | Improved |
| use primitive planes instead of BPSS supervoxels (ablation) | PPTr · Synthia 4D | mIoU: 83.43 -> 84.49 | Improved |
| use primitive planes instead of BPSS supervoxels (ablation) | PPTr · MSR-Action3D | accuracy: 91.98 -> 92.33 | Improved |
| vary primitive number from 100 to 400 (ablation) | PPTr · Synthia 4D | mIoU: 84.41, 84.49, 84.28 | No clear effect |
| use 10 primitives instead of 200 (ablation) | PPTr · Synthia 4D | mIoU: 84.49 -> 82.98 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PSTNet — PSTNet: Point Spatio-Temporal Convolution on Point Cloud Sequences (ICLR 2021) · better on 2 of 2 shared results
- MeteorNet — MeteorNet: Deep Learning on Dynamic 3D Point Cloud Sequences (ICCV 2019) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- X4D-SceneFormer: Enhanced Scene Understanding on 4D Point Cloud Videos through Cross-Modal Knowledge Transfer (AAAI 2024 · similarity 0.90)
- Adapting Pre-trained 3D Models for Point Cloud Video Understanding via Cross-frame Spatio-temporal Perception (CVPR 2025 · similarity 0.87)
- PvNeXt: Rethinking Network Design and Temporal Motion for Point Cloud Video Recognition (ICLR 2025 · similarity 0.86)
- Point Transformer V3: Simpler Faster Stronger (CVPR 2024 · similarity 0.86)
- Mamba4D: Efficient 4D Point Cloud Video Understanding with Disentangled Spatial-Temporal State Space Models (CVPR 2025 · similarity 0.85)
- Scene Flow Prior Based Point Cloud Completion with Masked Transformer (Student Abstract) (AAAI 2024 · similarity 0.85)
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.
