Research map / Detection and segmentation / Open-set and few-shot detection
Unsupervised Learning of Object Landmarks via Self-Training Correspondence
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| learn correspondence via clustering instead of affine-augmented pairs (equivariance) (ablation) | Ours · CelebA | NMI: 0.290 -> 0.620 | Helped |
| learn correspondence via clustering instead of affine-augmented pairs (equivariance) (ablation) | Ours · AFLW | NMI: 0.352 -> 0.635 | Helped |
| initialize with SuperPoint instead of R2D2 (ablation) | Ours · CelebA | forward error: 4.57 -> 4.12 | Helped |
| initialize with SuperPoint instead of R2D2 (ablation) | Ours · AFLW | forward error: 8.87 -> 7.37 | Helped |
| use over-segmented clusters M larger than number of true landmarks (best M=30 for 15 landmarks) (ablation) | Ours · AFLW synthetic keypoints | NMI | Helped |
| include 40% ground-truth points among initial keypoints (ablation) | Ours · AFLW synthetic keypoints | NMI / F-measure | Helped |
| proposed method vs prior landmark discovery methods | Ours · BBCPose | avg accuracy (%): 42.69 -> 61.18 | Helped |
| proposed method vs prior methods on MAFL/AFLW forward error with all training images | Ours · MAFL | forward NME (%): 3.19 -> 4.12 | Hurt |
Closest papers by meaning
- Object landmark discovery through unsupervised adaptation (NeurIPS 2019 · similarity 0.94)
- Unsupervised Learning of Object Landmarks through Conditional Image Generation (NeurIPS 2018 · similarity 0.91)
- Pose-Guided Self-Training with Two-Stage Clustering for Unsupervised Landmark Discovery (CVPR 2024 · similarity 0.90)
- Unsupervised 3D Structure Inference from Category-Specific Image Collections (CVPR 2024 · similarity 0.86)
- Jamais Vu: Exposing the Generalization Gap in Supervised Semantic Correspondence (NeurIPS 2025 · similarity 0.85)
- SCE-MAE: Selective Correspondence Enhancement with Masked Autoencoder for Self-Supervised Landmark Estimation (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.