Research map / Detection and segmentation / Tracking
Context-PIPs: Persistent Independent Particles Demands Spatial Context Features
Weikang BIAN, Zhaoyang Huang, Xiaoyu Shi, Yitong Dong, Yijin Li, Hongsheng Li
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 SOFE module (ablation) | PIPs · FlyingThings++ | ATE-Vis: 7.40 -> 6.91 | Improved |
| add SOFE module (ablation) | PIPs · FlyingThings++ | ATE-Occ: 24.4 -> 22.64 | Improved |
| add TAFA with cross-attention on top of SOFE (ablation) | Context-PIPs · FlyingThings++ | ATE-Vis: 6.91 -> 6.60 | Improved |
| predict feature weights instead of cross-attention in TAFA (ablation) | Context-PIPs · FlyingThings++ | ATE-Vis: 6.60 -> 7.15 | Hurt |
| increase PIPs MLP-Mixer depth from 12 to 16 (ablation) | PIPs · FlyingThings++ | ATE-Vis: 7.70 -> 7.69 | No clear effect |
| reduce MLP-Mixer depth of Context-PIPs to 3 (ablation) | Context-PIPs · FlyingThings++ | ATE-Vis: 7.70 -> 7.37 | Improved |
| increase PIPs MLP-Mixer depth from 6 to 12 (ablation) | PIPs · FlyingThings++ | ATE-Vis: 8.00 -> 7.70 | Improved |
| increase correlation radius r_c from 2 to 4 (ablation) | SOFE · FlyingThings++ | ATE-Vis: 13.02 -> 13.75 | Hurt |
| increase correlation radius r_c from 1 to 2 (ablation) | SOFE · FlyingThings++ | ATE-Vis: 13.60 -> 13.02 | Improved |
| increase predicted samples M from 1 to 9 (ablation) | SOFE · FlyingThings++ | ATE-Vis: 15.00 -> 12.18 | Improved |
| increase predicted samples M from 9 to 12 (ablation) | SOFE · FlyingThings++ | ATE-Vis: 12.18 -> 12.87 | Hurt |
| use M=1 sample in SOFE instead of PIPs baseline (ablation) | PIPs · FlyingThings++ | ATE-Vis: 14.42 -> 15.00 | Hurt |
Closest papers by meaning
- TrackIME: Enhanced Video Point Tracking via Instance Motion Estimation (NeurIPS 2024 · similarity 0.86)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · similarity 0.86)
- TAPIP3D: Tracking Any Point in Persistent 3D Geometry (NeurIPS 2025 · similarity 0.86)
- TAPVid-3D: A Benchmark for Tracking Any Point in 3D (NeurIPS 2024 · similarity 0.86)
- ETAP: Event-based Tracking of Any Point (CVPR 2025 · similarity 0.86)
- DriveTrack: A Benchmark for Long-Range Point Tracking in Real-World Videos (CVPR 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.
