Research map / 3D vision / Depth and stereo
HSCNet++: Hierarchical Scene Coordinate Classification and Regression for Visual Localization with Transformer
Shuzhe Wang, Zakaria Laskar, Iaroslav Melekhov, Xiaotian Li, Yi Zhao, Giorgos Tolias, Juho Kannala
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 transformer conditioning (HSCNet -> HSCNet++) | HSCNet · 7-Scenes | accuracy (%): 84.8 -> 88.7 | Improved |
| add transformer conditioning (HSCNet -> HSCNet++) | HSCNet · i7-Scenes | accuracy (%): 83.3 -> 88.3 | Improved |
| add FiLM conditioning mechanism (ablation) | HSCNet · i7-Scenes | accuracy (%): 70.3 -> 83.3 | Improved |
| add data augmentation (ablation) | HSCNet · i7-Scenes | accuracy (%): 71.5 -> 83.3 | Improved |
| add transformer encoders in classification branches (not only t_3D) (ablation) | HSCNet++ · i7-Scenes | accuracy (%): 84.9 -> 88.3 | Improved |
| increase MHA layers from 5 to 8 without classification-branch transformers (ablation) | HSCNet++-dagger · i7-Scenes | accuracy (%): 84.9 -> 84.3 | No clear effect |
| use conventional sine-cosine positional encoding instead of FiLM region encoding (ablation) | HSCNet++ · i7-Scenes | accuracy (%): 88.3 -> 65.5 | Hurt |
| add classification-branch transformers (ablation) | HSCNet++-dagger · i7-Scenes | sub-region accuracy (%): 68.7 -> 70.7 | Improved |
| add label propagation (ablation) | HSCNet++(S) · 7-Scenes sparse | accuracy (%): 62.0 -> 85.2 | Improved |
| add transformers (ablation) | HSCNet++(S) · 7-Scenes sparse | accuracy (%): 83.7 -> 85.2 | Improved |
| add symmetric cross-entropy loss (ablation) | HSCNet++(S) · 7-Scenes sparse | accuracy (%): 83.1 -> 85.2 | Improved |
| add reprojection loss (ablation) | HSCNet++(S) · 7-Scenes sparse | accuracy (%): 79.3 -> 85.2 | Improved |
Closest papers by meaning
- Hierarchical Scene Coordinate Classification and Regression for Visual Localization (CVPR 2020 · similarity 0.94)
- R-SCoRe: Revisiting Scene Coordinate Regression for Robust Large-Scale Visual Localization (CVPR 2025 · similarity 0.88)
- Deep Homography Estimation for Visual Place Recognition (AAAI 2024 · similarity 0.85)
- GLACE: Global Local Accelerated Coordinate Encoding (CVPR 2024 · similarity 0.85)
- HyperPose: Hypernetwork-Infused Camera Pose Localization and an Extended Cambridge Landmarks Dataset (CVPR 2025 · similarity 0.85)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · 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.
