Research map / Point clouds / Point cloud classification
PSTNet: Point Spatio-Temporal Convolution on Point Cloud Sequences
Hehe Fan, Xin Yu, Yuhang Ding, Yi Yang, Mohan Kankanhalli
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 PSTNet instead of MeteorNet | point-cloud action recognition · MSR-Action3D, 24 frames | accuracy: 88.50 -> 91.20 | Improved |
| use PSTNet instead of 3DV-PointNet++ | 3D action recognition · NTU RGB+D 120, Subject | accuracy: 82.4 -> 87.0 | Improved |
| use PSTNet instead of MeteorNet | 4D semantic segmentation · Synthia 4D | mIoU: 81.80 -> 82.24 | Improved |
| increase temporal kernel size from l=1 to l=3 (ablation) | PSTNet · Synthia 4D | mIoU: 80.79 -> 82.24; 1.45% | Improved |
| add point tracking across frames (ablation) | PSTNet · Moving MNIST Point Cloud | accuracy: 97.15 -> 98.74 | Improved |
| replace lightweight spatial convolution with PointConv (ablation) | PSTNet · MSR-Action3D, 16 frames | accuracy: 89.90 -> 90.24 | Improved |
| replace lightweight spatial convolution with PointConv (ablation) | PSTNet · MSR-Action3D, 16 frames | running time per clip: 31.92 -> 102.72 | Hurt |
| remove the sharing kernel (ablation) | PSTNet spatial convolution · MSR-Action3D, 16 frames | accuracy: 89.90 -> 58.25 | Hurt |
| remove the sharing kernel (ablation) | PSTNet spatial convolution · MSR-Action3D, 24 frames | accuracy: 91.20 -> 61.67 | Hurt |
| remove the displacement kernel (ablation) | PSTNet spatial convolution · MSR-Action3D, 16 frames | accuracy: 89.90 -> 86.35 | Hurt |
| remove the displacement kernel (ablation) | PSTNet spatial convolution · MSR-Action3D, 24 frames | accuracy: 91.20 -> 87.46 | Hurt |
| disentangle spatial and temporal modeling instead of combining space and time in one kernel (ablation) | PSTNet · MSR-Action3D | accuracy | Improved |
Later papers that compared with this method
9 later accepted papers report PSTNet as a baseline in their results tables; 7 of them report a better number on most of the results they share with it.
- Paving the Way for Point Cloud Video Representation Learning Using a PDE Model (TPAMI 2026 · better on 5 of 5 shared results)
- UST-SSM: Unified Spatio-Temporal State Space Models for Point Cloud Video Modeling (ICCV 2025 · better on 3 of 3 shared results)
- PvNeXt: Rethinking Network Design and Temporal Motion for Point Cloud Video Recognition (ICLR 2025 · better on 0 of 1 shared result)
- Dynamic 3D Point Cloud Sequences as 2D Videos (TPAMI 2024 · better on 3 of 3 shared results)
- Contrastive Predictive Autoencoders for Dynamic Point Cloud Self-Supervised Learning (AAAI 2023 · better on 1 of 2 shared results)
- PointCMP: Contrastive Mask Prediction for Self-Supervised Learning on Point Cloud Videos (CVPR 2023 · better on 1 of 1 shared result)
- Point Contrastive Prediction with Semantic Clustering for Self-Supervised Learning on Point Cloud Videos (ICCV 2023 · better on 1 of 1 shared result)
- No Pain, Big Gain: Classify Dynamic Point Cloud Sequences With Static Models by Fitting Feature-Level Space-Time Surfaces (CVPR 2022 · better on 4 of 5 shared results)
1 more not listed.
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MeteorNet — MeteorNet: Deep Learning on Dynamic 3D Point Cloud Sequences (ICCV 2019) · better on 8 of 8 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
- Point Spatio-Temporal Transformer Networks for Point Cloud Video Modeling (TPAMI 2022 · similarity 0.93)
- Deep Hierarchical Representation of Point Cloud Videos via Spatio-Temporal Decomposition (TPAMI 2021 · similarity 0.90)
- UST-SSM: Unified Spatio-Temporal State Space Models for Point Cloud Video Modeling (ICCV 2025 · similarity 0.89)
- Dynamic 3D Point Cloud Sequences as 2D Videos (TPAMI 2024 · similarity 0.89)
- Point 4D Transformer Networks for Spatio-Temporal Modeling in Point Cloud Videos (CVPR 2021 · similarity 0.89)
- 3DInAction: Understanding Human Actions in 3D Point Clouds (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.
