How deepscenario/ortholoc works
OrthoLoC is a UAV localization and calibration framework built around aligned orthophotos, DSMs, query imagery, and camera ground truth. The principal workflow loads dataset samples, matches query-to-geodata images, converts matches into 2D–3D constraints, then estimates camera pose and optionally intrinsics. Supporting branches provide ground-truth matching, benchmarking, dataset downloading, and visualization. The graph reflects sampled runtime modules plus README-documented entry points; script internals and unsampled benchmark/visualization calls are therefore shown conservatively.
Components 18
Data Access
- Dataset Loader
ortholoc/dataset/OrthoLoC.py - Data IO
ortholoc/utils/io.py - Geodata Samples
ortholoc/assets/demo
Image Matching
- Matching CLI
ortholoc/scripts/run_matching.py - IMCUI Matcher
ortholoc/image_matching/MatcherIMCUI.py - Ground Truth Matcher
ortholoc/image_matching/MatcherGT.py - Correspondence Model
ortholoc/correspondences/Correspondences.py - 2D-2D Matches
ortholoc/correspondences/Correspondences2D2D.py
Localization
- Localization CLI
ortholoc/scripts/run_localization.py - 2D-3D Constraints
ortholoc/correspondences/Correspondences2D3D.py - Geometry Utilities
ortholoc/utils/geometry.py - Pose Solver
ortholoc/utils/pose.py - Camera Result
ortholoc/correspondences/Correspondences2D3D.py
Evaluation Outputs
- Benchmark CLI
ortholoc/scripts/run_benchmark.py - Localization Metrics
ortholoc/utils/metrics.py - Visualization Tools
ortholoc/utils/plot.py
Other components
- Researcher
- Dataset Server
Connections 20
- Researcher → Matching CLI: runs matching
- Researcher → Localization CLI: runs localization
- Researcher → Benchmark CLI: runs benchmark
- Dataset Loader → Data IO: loads samples
- Data IO → Geodata Samples: reads files
- Data IO → Dataset Server: downloads data
- Matching CLI → Dataset Loader: loads dataset
- Localization CLI → Dataset Loader: loads dataset
- IMCUI Matcher → 2D-2D Matches: builds matches
- Ground Truth Matcher → Dataset Loader: reads ground truth
- Ground Truth Matcher → 2D-2D Matches: returns matches
- 2D-2D Matches → Correspondence Model: uses correspondence model
- 2D-2D Matches → Geometry Utilities: normalizes points
- 2D-2D Matches → 2D-3D Constraints: converts matches
- Localization CLI → IMCUI Matcher: matches imagery
- Localization CLI → 2D-3D Constraints: localizes camera
- 2D-3D Constraints → Pose Solver: estimates pose
- Pose Solver → Camera Result: returns camera
- Benchmark CLI → Localization Metrics: evaluates results
- Visualization Tools → Geometry Utilities: plots projections