Research map / Vision-language and multimodal / Cross-modal learning
SCPNet: Unsupervised Cross-modal Homography Estimation via Intra-modal Self-supervised Learning
Runmin Zhang, Jun Ma, Lun Luo, Beinan Yu, Shu-Jie Chen, Junwei Li, Hui-Liang Shen, Si-Yuan Cao
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 intra-modal self-supervised learning (versus cross-modal intensity loss only) (ablation) | SCPNet · GoogleMap | MACE: 24.64 -> 10.01 (with projection) | Improved |
| add correlation to the self-supervised network (ablation) | SCPNet · GoogleMap | MACE: 13.06 -> 9.68 | Improved |
| add consistent feature map projection to the self-supervised network (ablation) | SCPNet · GoogleMap | MACE: 13.06 -> 10.01 | Improved |
| add cross-modal intensity-based learning on top of self-supervised, correlation and projection (ablation) | SCPNet · GoogleMap | MACE: 7.70 -> 4.35 | Improved |
| add correlation to cross-modal-only training with projection (ablation) | SCPNet · GoogleMap | MACE: 24.64 -> 24.80 | No clear effect |
| SCPNet vs supervised MHN on GoogleMap | SCPNet · GoogleMap | MACE: 5.06 -> 4.35 | Improved |
| SCPNet vs stronger supervised methods (LocalTrans, IHN, RHWF) | SCPNet · Flash/no-flash | MACE: 2.67 vs IHN 0.80 | Hurt |
| SCPNet vs supervised MHN on Harvard | SCPNet · Harvard | MACE: 5.35 -> 4.00 | Improved |
| add two-branch intra-modal self-supervised learning (ablation) | SCPNet · Training cost | training time (hours): 3.18 -> 6.98 hours; memory 4622 -> 9144 MB | Hurt |
Closest papers by meaning
- SSHNet: Unsupervised Cross-modal Homography Estimation via Problem Reformulation and Split Optimization (CVPR 2025 · similarity 0.93)
- CrossHomo: Cross-Modality and Cross-Resolution Homography Estimation (TPAMI 2024 · similarity 0.90)
- HomoDiff: Noise-Robust Multi-modal Homography Estimation via Diffusion Model (TPAMI 2026 · similarity 0.89)
- MCNet: Rethinking the Core Ingredients for Accurate and Efficient Homography Estimation (CVPR 2024 · similarity 0.88)
- Unsupervised Homography Estimation on Multimodal Image Pair via Alternating Optimization (NeurIPS 2024 · similarity 0.87)
- Adapting Dense Matching for Homography Estimation with Grid-based Acceleration (CVPR 2025 · 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.
