Research map / Detection and segmentation / Open-set and few-shot detection
Object landmark discovery through unsupervised adaptation
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 |
|---|---|---|---|
| adapt a frozen supervised core through learned projection matrices instead of training from scratch | object landmark detection · MAFL | forward MSE: 5.00 -> 3.99 | Helped |
| adapt a frozen supervised core through learned projection matrices instead of fine-tuning | object landmark detection · MAFL | backward MSE: 4.43 -> 3.55 | Helped |
| adapt a frozen supervised core through learned projection matrices instead of training from scratch | object landmark detection · AFLW | forward MSE: 7.65 -> 6.69 | Helped |
| adapt a frozen supervised core through learned projection matrices instead of fine-tuning | object landmark detection · AFLW | backward MSE: 11.14 -> 9.19 | Helped |
| adapt a frozen supervised core through learned projection matrices instead of fine-tuning | object landmark detection · LS3D | forward MSE: 24.79 -> 26.41 | Hurt |
| adapt a frozen supervised core through learned projection matrices instead of fine-tuning | object landmark detection · LS3D | backward MSE: 5.81 -> 5.44 | Helped |
| adapt a frozen supervised face-landmark core to discover body landmarks | landmark detection on BBC-Pose · BBC-Test | consistency error: 7.68 -> 5.54 | Helped |
| use a 1-epoch rather than 110-epoch core network (ablation) | proposed adaptation · AFLW | consistency error (Avg.): 2.52 -> 15.09 | Hurt |
| use a 5-epoch rather than 110-epoch core network (ablation) | proposed adaptation · AFLW | consistency error (Avg.): 2.52 -> 14.32 | Hurt |
| use a 10-epoch rather than 110-epoch core network (ablation) | proposed adaptation · AFLW | consistency error (Avg.): 2.52 -> 5.30 | Hurt |
| train target adaptation with 10 images rather than all images (ablation) | proposed adaptation · AFLW | consistency error (Avg.): 2.52 -> 3.11 | Helped |
| train target adaptation with 100 images rather than all images (ablation) | proposed adaptation · AFLW | consistency error (Avg.): 2.52 -> 3.44 | Hurt |
Closest papers by meaning
- Unsupervised Learning of Object Landmarks via Self-Training Correspondence (NeurIPS 2020 · similarity 0.94)
- Unsupervised Discovery of Object Landmarks as Structural Representations (CVPR 2018 · similarity 0.90)
- Unsupervised Learning of Object Landmarks through Conditional Image Generation (NeurIPS 2018 · similarity 0.89)
- Pose-Guided Self-Training with Two-Stage Clustering for Unsupervised Landmark Discovery (CVPR 2024 · similarity 0.87)
- unMORE: Unsupervised Multi-Object Segmentation via Center-Boundary Reasoning (ICML 2025 · similarity 0.86)
- Unsupervised 3D Structure Inference from Category-Specific Image Collections (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.