Research map / 3D vision / 3D generation
SNAP: Self-Supervised Neural Maps for Visual Positioning and Semantic Understanding
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 |
|---|---|---|---|
| replace SfM+SuperGlue with SNAP-large multi-modal | single-image positioning · All (100%) | AUC @2.5m/5deg: 39.2 -> 44.4 | Helped |
| use SfM+SuperGlue instead of SNAP on easy queries | structure-based · Easy (25%) | AUC @2.5m/5deg: 63.0 -> 48.9 | Hurt |
| SNAP-large vs OrienterNet on hard queries | SNAP · Hard (25%) | AUC @2.5m/5deg: 24.8 -> 34.5 | Helped |
| fuse StreetView and aerial instead of StreetView only | SNAP-large · All (100%) | AUC @2.5m/5deg: 41.0 -> 44.4 | Helped |
| use aerial only instead of StreetView only | SNAP-large · All (100%) | AUC @2.5m/5deg: 41.0 -> 24.8 | Hurt |
| scale backbone from ResNet-50 to ResNet-152x2 | SNAP multi-modal · All (100%) | AUC @2.5m/5deg: 39.9 -> 44.4 | Helped |
| fixed-height plane instead of 3D lifting (ablation) | SNAP StreetView-only · All (100%) | AUC @2.5m/5deg: 34.0 -> 28.2 | Hurt |
| remove monocular occupancy weighting (ablation) | SNAP StreetView-only · All (100%) | AUC @2.5m/5deg: 34.0 -> 24.7 | Hurt |
| remove multi-view variance and occupancy (ablation) | SNAP StreetView-only · All (100%) | AUC @2.5m/5deg: 34.0 -> 22.7 | Hurt |
| replace occupancy volume with ray-conditioning MLP (ablation) | SNAP StreetView-only · All (100%) | AUC @2.5m/5deg: 34.0 -> 08.6 | Hurt |
| average instead of max vertical pooling (ablation) | SNAP StreetView-only · All (100%) | AUC @2.5m/5deg: 34.0 -> 27.3 | Hurt |
| average instead of max in multi-modal fusion (ablation) | SNAP · All (100%) | AUC @2.5m/5deg: 39.3 -> 39.0 | No clear effect |
Closest papers by meaning
- Geometry-Aware Learning of Maps for Camera Localization (CVPR 2018 · similarity 0.89)
- SNI-SLAM: Semantic Neural Implicit SLAM (CVPR 2024 · similarity 0.86)
- MAPLM: A Real-World Large-Scale Vision-Language Benchmark for Map and Traffic Scene Understanding (CVPR 2024 · similarity 0.86)
- Less Decoder is More Encoder: Geometric Representation Learning from Novel View Synthesis (NeurIPS 2026 · similarity 0.85)
- Textual Supervision Enhances Geospatial Representations in Vision-Language Models (ICML 2026 · similarity 0.85)
- Swift-Mapping: Online Neural Implicit Dense Mapping in Urban Scenes (AAAI 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.