Research map / Detection and segmentation / CNN features
D2-Net: A Trainable CNN for Joint Description and Detection of Local Features
Check 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 |
|---|---|---|---|
| Use D2 MS instead of DenseSfM dense matching | Aachen Day-Night localization · Aachen Day-Night | localized queries (%) at 0.5m, 2°: 39.8 -> 43.9 | Helped |
| Use D2 MS instead of DenseSfM dense matching | Aachen Day-Night localization · Aachen Day-Night | localized queries (%) at 1.0m, 5°: 60.2 -> 67.3 | Helped |
| Add multiscale detection rather than using single-scale detection (ablation) | D2-Net · Aachen Day-Night | localized queries (%) at 0.5m, 2°: 41.8 -> 43.9 | Helped |
| Fine-tune on MegaDepth (ablation) | D2-Net single-scale · Aachen Day-Night | localized queries (%) at 0.5m, 2°: 41.8 -> 44.9 | Helped |
| Fine-tune on MegaDepth (ablation) | D2-Net multiscale · Aachen Day-Night | localized queries (%) at 1.0m, 5°: 67.3 -> 64.3 | Hurt |
| Use D2 MS rather than affine RootSIFT features | Direct PE localization · InLoc | localized queries (%) within 0.5m: 26.4 -> 40.4 | Helped |
| Use D2 MS rather than dense PE matching | Sparse PE localization · InLoc | localized queries (%) within 0.5m: 46.2 -> 48.6 | Helped |
| Combine D2 MS with Dense PV pose verification | Sparse PE localization · InLoc | localized queries (%) within 0.5m: 48.6 -> 54.1 | Helped |
| Combine D2 MS with InLoc pose selection | InLoc localization · InLoc | localized queries (%) within 0.5m: 56.5 -> 61.1 | Helped |
Closest papers by meaning
- LoFTR: Detector-Free Local Feature Matching With Transformers (CVPR 2021 · similarity 0.86)
- Neighbourhood Consensus Networks (NeurIPS 2018 · similarity 0.86)
- Hierarchical Scene Coordinate Classification and Regression for Visual Localization (CVPR 2020 · similarity 0.86)
- ASLFeat: Learning Local Features of Accurate Shape and Localization (CVPR 2020 · similarity 0.86)
- LocNet: Improving Localization Accuracy for Object Detection (CVPR 2016 · similarity 0.85)
- Rooms from Motion: Un-posed Indoor 3D Object Detection as Localization and Mapping (NeurIPS 2025 · similarity 0.84)
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.