Research map / Label-efficient and robust learning / Semi-supervised learning
Semi-Supervised Keypoint Detector and Descriptor for Retinal Image Matching
Jiazhen Liu, Xirong Li, Qijie Wei, Jie Xu, Dayong Ding
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 progressive keypoint expansion (PKE) (ablation) | SuperRetina · FIRE | mAUC: 0.685 -> 0.755 | Improved |
| add progressive keypoint expansion (PKE) (ablation) | SuperRetina · CLINICAL | EER (%): 5.14 -> 0.83 | Improved |
| add content-based matching to PKE (double matching) (ablation) | SuperRetina · FIRE | mAUC: 0.670 -> 0.755 | Improved |
| use hardest plus random negative instead of semi-hard negative sampling (ablation) | SuperRetina description loss · FIRE | mAUC: 0.407 -> 0.755 | Improved |
| add descriptor upsampling (ablation) | SuperRetina · FIRE | mAUC: 0.697 -> 0.755 | Improved |
| add 2D Gaussian blur (ablation) | SuperRetina · FIRE | mAUC: 0.574 -> 0.755 | Improved |
| use Dice loss instead of cross-entropy (ablation) | SuperRetina · FIRE | mAUC: 0.653 -> 0.755 | Improved |
| use Dice loss instead of weighted cross-entropy (ablation) | SuperRetina · FIRE | mAUC: 0.704 -> 0.755 | Improved |
| use SuperRetina detector instead of SIFT detector (ablation) | SuperRetina · FIRE | mAUC: 0.585 -> 0.755 | Improved |
| use SuperRetina descriptor instead of RootSIFT (ablation) | SuperRetina · FIRE | mAUC: 0.705 -> 0.755 | Improved |
| use PBO-detected initial keypoints instead of manual labels (ablation) | SuperRetina · FIRE | mAUC: 0.755 -> 0.750 | No clear effect |
| finetune on retinal training data | SuperPoint · FIRE | mAUC: 0.674 -> 0.661 | Hurt |
Closest papers by meaning
- SuperJunction: Learning-Based Junction Detection for Retinal Image Registration (AAAI 2024 · similarity 0.91)
- R2D2: Reliable and Repeatable Detector and Descriptor (NeurIPS 2019 · similarity 0.85)
- Dense Keypoints via Multiview Supervision (NeurIPS 2021 · similarity 0.84)
- FOCUS: Benchmarking Retinal Model Generalization from Foundation Vision Encoders to Multimodal LLMs (NeurIPS 2026 · similarity 0.83)
- ESCAPE: Encoding Super-keypoints for Category-Agnostic Pose Estimation (CVPR 2024 · similarity 0.82)
- Towards Complementary Keypoint Detection via Mixture of Detectors (NeurIPS 2026 · similarity 0.82)
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.
