Research map / Point clouds / Segmentation and detection
CPR++: Object Localization via Single Coarse Point Supervision
Xuehui Yu, Pengfei Chen, Kuiran Wang, Xumeng Han, Guorong Li, Zhenjun Han, Qixiang Ye, Jianbin Jiao
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 MIL loss (vs pos loss treating all bag points as positive) (ablation) | CPR · COCO val | mAP@1.0: 42.72 vs 55.46 | Improved |
| add MIL loss on top of annotation and negative loss (ablation) | CPR · COCO val | mAP@1.0: 51.82 vs 55.46 | Improved |
| add annotation loss (ablation) | CPR · COCO val | mAP@1.0: 54.24 vs 55.46 | Improved |
| add negative loss (ablation) | CPR · COCO val | mAP@1.0: 39.45 -> 55.46 | Improved |
| increase sampling radius R from 5 to 8 (ablation) | CPR · COCO val | mAP@1.0: 53.32 -> 55.46 | Improved |
| increase sampling radius R beyond 25 (ablation) | CPR · COCO val | mAP@1.0: 55.46 -> 53.85 | Hurt |
| replace circle sampling with rectangles (ablation) | CPR · COCO val | mAP@1.0: 55.46 vs 55.39 | No clear effect |
| vary sample density u0 from 4 to 12 (ablation) | CPR · COCO val | mAP@1.0: 55.25 to 55.46 | No clear effect |
| iterate CPR independently with fixed radius (iter 1 to 6) (ablation) | CPR · COCO val | mAP@1.0: 55.46 -> 55.13 | No clear effect |
| use dynamic radius cascade (Cascade CPR II) instead of fixed-radius cascade (Cascade CPR I), no var (ablation) | CPR++ · COCO val | mAP@1.0: 55.36 -> 57.57 | Improved |
| add variance regularization to Cascade CPR II (ablation) | CPR++ · COCO val | mAP@1.0: 57.57 -> 58.22 | Improved |
| increase number of stages K_s from 1 to 2 (ablation) | CPR++ · COCO val | mAP@1.0: 54.93 -> 57.57 | Improved |
Closest papers by meaning
- Object Localization Under Single Coarse Point Supervision (CVPR 2022 · similarity 0.95)
- Unsupervised Object Localization with Representer Point Selection (ICCV 2023 · similarity 0.88)
- RepPoints: Point Set Representation for Object Detection (ICCV 2019 · similarity 0.88)
- Leveraging Imagery Data with Spatial Point Prior for Weakly Semi-supervised 3D Object Detection (AAAI 2024 · similarity 0.86)
- LiSA: LiDAR Localization with Semantic Awareness (CVPR 2024 · similarity 0.85)
- DiffuBox: Refining 3D Object Detection with Point Diffusion (NeurIPS 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.
