Research map / Label-efficient and robust learning / Recognition and long-tail
Cross-Descriptor Visual Localization and Mapping
Mihai Dusmanu, Ondrej Miksik, Johannes L. Schönberger, Marc Pollefeys
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 |
|---|---|---|---|
| translate SIFT map descriptors to HardNet instead of keeping SIFT | SIFT map · Aachen day, 0.5m 5deg | % localized queries: 88.7 -> 90.9 | Improved |
| translate BRIEF map descriptors to HardNet instead of keeping BRIEF | BRIEF map · Aachen night, 0.5m 5deg | % localized queries: 36.7 -> 49.0 | Improved |
| translate SIFT map to BRIEF queries instead of native SIFT | SIFT map · Aachen night, 0.5m 5deg | % localized queries: 61.2 -> 23.5 | Hurt |
| translate HardNet descriptors to SIFT/BRIEF in continuous deployment | HardNet map · Aachen night, 0.5m 5deg | % localized queries: 72.4 -> 26.5 | Hurt |
| use the joint embedding for heterogeneous map instead of native BRIEF queries | BRIEF queries · Aachen day, 0.5m 5deg | % localized queries: 81.4 -> 86.7 | Improved |
| use joint-embedding collaborative map instead of native HardNet | HardNet · Aachen night, 0.5m 5deg | % localized queries: 72.4 -> 61.2 | Hurt |
| use Embed instead of per-quarter native SOSNet maps | collaborative mapping · Madrid Metropolis, infinite threshold | % localized images: 21.2 -> 92.1 | Improved |
| use Embed instead of Progressive translation | collaborative mapping · Tower of London, infinite threshold | % localized images: 96.6 -> 97.9 | Improved |
| increase joint embedding dimension from 128 to 256 (ablation) | Embed · Aachen night, 0.5m 5deg | % localized queries: 60.2 -> 64.3 | No clear effect |
| reduce joint embedding dimension from 128 to 32 (ablation) | Embed · Aachen night, 0.5m 5deg | % localized queries: 60.2 -> 49.0 | Hurt |
| reduce joint embedding dimension from 128 to 16 (ablation) | Embed · Aachen night, 0.5m 5deg | % localized queries: 60.2 -> 33.7 | Hurt |
| add matching loss to the encoder-decoder (ablation) | encoder-decoder · HPatches sequences | matching accuracy | Improved |
Closest papers by meaning
- Towards Visual Localization Interoperability: Cross-Feature for Collaborative Visual Localization and Mapping (ICCV 2025 · similarity 0.93)
- Visual Cross-View Metric Localization with Dense Uncertainty Estimates (ECCV 2022 · similarity 0.89)
- From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · similarity 0.87)
- CroCoDL: Cross-device Collaborative Dataset for Localization (CVPR 2025 · similarity 0.87)
- Enhancing Visual Localization with Cross-Domain Image Generation (ICML 2025 · similarity 0.86)
- Aligning Geometric Spatial Layout in Cross-View Geo-Localization via Feature Recombination (AAAI 2024 · similarity 0.86)
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.
