Research map / Label-efficient and robust learning / Self-supervised and contrastive
GECO: Geometrically Consistent Embedding with Lightspeed Inference
Regine Hartwig, Dominik Muhle, Riccardo Marin, Daniel Cremers
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 |
|---|---|---|---|
| replace argmax-based supervision with optimal transport soft assignment loss with dustbin | DINOv2-B + LoRA · PFPascal | PCK@0.1: 86.0 -> 92.1 (DINOv2-B vs GECO) | Improved |
| use lightweight DINOv2 plus LoRA instead of diffusion-feature based Geo | GECO · APK inference time | time (ms): 2127 -> 40 | Improved |
| use GECO rather than Geo on SPair | GECO · SPair test | PCK@0.1: 90.1 -> 85.2 (Geo vs GECO) | Hurt |
| use GECO on CUB versus the DINOv2-B backbone | GECO · CUB (unseen in training) | PCK@0.1: 92.8 -> 92.5 | No clear effect |
| scale backbone from DINOv2-S to DINOv2-B (ablation) | GECO · SPair test | PCK@0.1: 78.0 -> 85.2 | Improved |
| use mask-based marginals instead of standard SuperGlue marginals (ablation) | GECO · CUB (unseen in training) | PCK@0.1: 89.5 -> 92.5 | Improved |
| use KL-regularized soft marginals instead of hard marginal constraints (ablation) | GECO · PascalParts segmentation | mean mIoU: 36.4 -> 37.9 | Improved |
| sample fewer image pairs per category (K 100-400) when jointly training (ablation) | GECO · PFPascal, SPair, CUB | PCK | Hurt |
| apply GECO to preserve pretrained features versus Geo | GECO · CUB (unseen in training) | PCK@0.1: 88.4 -> 92.5 | Improved |
| use GECO features for part segmentation versus DINOv2-B overall | GECO · PascalParts segmentation | mean mIoU: 39.8 -> 37.9 | Hurt |
Later papers that compared with this method
1 later accepted paper reports GECO as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- MARCO: Navigating the Unseen Space of Semantic Correspondence (CVPR 2026 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- GeCo: Geometry-Consistent Regularization for Domain Generalized Semantic Segmentation (CVPR 2026 · similarity 0.88)
- SemLight: Distilled Semantic–Geometric Fusion for Efficient Local Feature Matching (ECCV 2026 · similarity 0.86)
- SGMatch: Semantic-Guided Non-Rigid Shape Matching with Flow Regularization (ECCV 2026 · similarity 0.86)
- Self-Supervised Learning of Geometrically Stable Features Through Probabilistic Introspection (CVPR 2018 · similarity 0.86)
- G$^2$VLM: Geometry Grounded Vision Language Model with Unified 3D Reconstruction and Spatial Reasoning (CVPR 2026 · similarity 0.85)
- Beyond 3D VQAs: Injecting 3D Spatial Priors into Vision-Language Models for Enhanced Geometric Reasoning (CVPR 2026 · 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.
