Research map / Point clouds / Point cloud classification
FastPCI: Motion-Structure Guided Fast Point Cloud Frame Interpolation
tianyu zhang, Guocheng Qian, Jin Xie, Jian Yang
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 structure branch (structure-aware motion estimation) (ablation) | FastPCI · KITTI odometry | EMD: 67.60 -> 53.78 | Improved |
| use Transformer with cross-attention instead of convolution (ablation) | FastPCI · KITTI odometry | EMD: 56.38 -> 53.78 | Improved |
| use dual-direction cross-attention instead of self-attention on forward features (ablation) | FastPCI · Nuscenes | EMD: 181.58 -> 173.64 | Improved |
| add motion compensation block (ablation) | FastPCI · KITTI odometry | EMD: 55.29 -> 53.78 | Improved |
| add RefineNet (ablation) | FastPCI · Argoverse 2 sensor | EMD: 58.17 -> 57.95 | No clear effect |
| add forward half-way cycle loss L_CD1 (ablation) | FastPCI · KITTI odometry | EMD: 58.47 -> 53.78 | Improved |
| add backward warp loss L_CD2 (ablation) | FastPCI · KITTI odometry | EMD: 55.54 -> 53.78 | Improved |
| add pyramid multiscale loss L_ms (ablation) | FastPCI · Nuscenes | CD: 1.37 -> 1.11 | Improved |
| add pyramid multiscale and half-way cycle losses to the interpolation loss (ablation) | FastPCI · Argoverse 2 sensor | CD: 0.92 -> 0.73 | Improved |
| FastPCI vs NeuralPCI inference speed | FastPCI · Inference time, 8192 points | runtime: 60822ms -> 109ms | Improved |
| vary pyramid loss weights alpha in 0.025 to 0.25 | FastPCI · point cloud interpolation | CD/EMD | No clear effect |
Later papers that compared with this method
1 later accepted paper reports FastPCI as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- DiffPCI: Large Motion Point Cloud frame Interpolation with Diffusion Model (ICCV 2025 · better on 6 of 6 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- NeuralPCI — NeuralPCI: Spatio-Temporal Neural Field for 3D Point Cloud Multi-Frame Non-Linear Interpolation (CVPR 2023) · better on 7 of 7 shared results
- PointINet — PointINet: Point Cloud Frame Interpolation Network (AAAI 2021) · better on 7 of 7 shared results
- PointPWC-Net — PointPWC-Net: Cost Volume on Point Clouds for (Self-)Supervised Scene Flow Estimation (ECCV 2020) · better on 1 of 1 shared result
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
- NeuralPCI: Spatio-Temporal Neural Field for 3D Point Cloud Multi-Frame Non-Linear Interpolation (CVPR 2023 · similarity 0.89)
- DiffPCI: Large Motion Point Cloud frame Interpolation with Diffusion Model (ICCV 2025 · similarity 0.89)
- PointPWC-Net: Cost Volume on Point Clouds for (Self-)Supervised Scene Flow Estimation (ECCV 2020 · similarity 0.86)
- HPLFlowNet: Hierarchical Permutohedral Lattice FlowNet for Scene Flow Estimation on Large-Scale Point Clouds (CVPR 2019 · similarity 0.86)
- Interpolated Convolutional Networks for 3D Point Cloud Understanding (ICCV 2019 · similarity 0.86)
- SCTN: Sparse Convolution-Transformer Network for Scene Flow Estimation (AAAI 2022 · 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.
