Research map / Label-efficient and robust learning / Domain adaptation and generalization
You Share Beliefs, I Adapt: Progressive Heterogeneous Collaborative Perception
Hao Si, Ehsan Javanmardi, Manabu Tsukada
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 PHCP inference-time adapter self-training | direct fusion baseline · OPV2V-H heterogeneous test (16 scenarios, LiDAR LP/LS) | mSAP@0.7: 53.0 -> 85.9 | Improved |
| add PHCP adapter | direct fusion baseline · OPV2V-H heterogeneous test (16 scenarios, LiDAR LP/LS) | mSAP@0.3: 59.7 -> 92.9 | Improved |
| use only a single shot of unlabeled data for adapter training (ablation) | PHCP · OPV2V-H heterogeneous test (16 scenarios, LiDAR LP/LS) | AP@IoU 0.7: approximately 50% | Improved |
| increase the number of shots (ablation) | PHCP · OPV2V-H heterogeneous test (16 scenarios, LiDAR LP/LS) | AP@IoU 0.7 | Improved |
| raise pseudo-label confidence threshold from 0.2 to 0.5 (ablation) | PHCP · OPV2V-H heterogeneous test (16 scenarios, LiDAR LP/LS) | mSAP@0.7: 85.0 -> 85.9 | Improved |
| raise pseudo-label confidence threshold from 0.5 to 0.7 (ablation) | PHCP · OPV2V-H heterogeneous test (16 scenarios, LiDAR LP/LS) | mSAP@0.7: 85.9 -> 85.8 | No clear effect |
| use soft confidence labels instead of thresholded labels (ablation) | PHCP · OPV2V-H heterogeneous test (16 scenarios, LiDAR LP/LS) | wSAP@0.7: 68.0 -> 67.0 | No clear effect |
| adapt heterogeneous pair with PHCP versus homogeneous PointPillar | PHCP · Heterogeneous vs homogeneous (OPV2V) | mSAP@0.7: 93.1 -> 85.9 | Hurt |
| PHCP versus HEAL trained on full labeled data | PHCP · OPV2V-H heterogeneous test (16 scenarios, LiDAR LP/LS) vs fully trained methods | mSAP@0.7: 91.7 -> 87.1 | Hurt |
| PHCP (few-shot unlabeled) versus V2X-ViT fully trained | V2X-ViT · OPV2V-H heterogeneous test (16 scenarios, LiDAR LP/LS) vs fully trained methods | mSAP@0.7: 82.8 -> 87.1 | Improved |
| increase shots from 1 to 5 (ablation) | PHCP · PHCP adapter training cost | training time (s): 1.49 -> 2.39 | Hurt |
Closest papers by meaning
- An Extensible Framework for Open Heterogeneous Collaborative Perception (ICLR 2024 · similarity 0.90)
- One is Plenty: A Polymorphic Feature Interpreter for Immutable Heterogeneous Collaborative Perception (CVPR 2025 · similarity 0.88)
- RCP-Bench: Benchmarking Robustness for Collaborative Perception Under Diverse Corruptions (CVPR 2025 · similarity 0.87)
- Transfer Your Perspective: Controllable 3D Generation from Any Viewpoint in a Driving Scene (CVPR 2025 · similarity 0.87)
- CoPEFT: Fast Adaptation Framework for Multi-Agent Collaborative Perception with Parameter-Efficient Fine-Tuning (AAAI 2025 · similarity 0.87)
- Pragmatic Heterogeneous Collaborative Perception via Generative Communication Mechanism (NeurIPS 2025 · similarity 0.87)
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.
