How brz1409/bathyfacto works
BathyFacto is a Nerfstudio plugin for training refraction-aware neural radiance fields from underwater imagery and exporting bathymetric point clouds. The user workflow is to obtain or build a mask- and metadata-rich dataset, train/evaluate the two-media model, then export points in the dataset’s global frame. The graph preserves the documented dataset-building, baseline, and download capabilities. Edges reflect explicit README claims or shown references and calls; unsampled implementation details are not inferred.
Main entry points: bathyfacto_model.py BathyFacto model · bathyfacto_config.py Training methods · export.py Point-cloud exporter · build_dataset.py Dataset builder
Components 19
Dataset preparation
- Dataset downloader
scripts/download_dataset.py - Metashape XML reader
bathyfacto/metashape_xml.py - Frame normalization
bathyfacto/normalization.py - Dataset builder
bathyfacto/build_dataset.py - BathyFacto dataset
Training and parsing
- Parser registration
bathyfacto/bathyfacto_dataparser_plugin.py - Dataset parser
bathyfacto/bathyfacto_dataparser.py - Masked ray data
bathyfacto/bathyfacto_datamanager.py - Training methods
bathyfacto/bathyfacto_config.py - Train and evaluate
bathyfacto/bathyfacto_pipeline.py
Model and geometry
- BathyFacto model
bathyfacto/bathyfacto_model.py - Radiance field
bathyfacto/bathyfacto_field.py - Two-media geometry
bathyfacto/two_media_geometry.py
Point-cloud export
- Point-cloud exporter
bathyfacto/export.py - Bathymetric point cloud
Other components
- Researcher
- Metashape project
- Source dataset
- Nerfstudio runtime
Connections 27
- Researcher → Dataset downloader: downloads
README.md - Source dataset → Dataset downloader: retrieves from
README.md - Researcher → Dataset builder: builds dataset
README.md - Metashape project → Dataset builder: provides project
README.md - Dataset builder → Metashape XML reader: parses XML
bathyfacto/build_dataset.py - Dataset builder → Frame normalization: uses transforms
bathyfacto/build_dataset.py - Dataset builder → BathyFacto dataset: writes dataset
README.md - Dataset downloader → BathyFacto dataset: provides dataset
README.md - BathyFacto dataset → Dataset parser: loads dataset
README.md - Parser registration → Dataset parser: registers parser
bathyfacto/bathyfacto_dataparser_plugin.py - Dataset parser → Masked ray data: supplies parsed data
- Training methods → Parser registration: configures parser
bathyfacto/bathyfacto_config.py - Training methods → Dataset parser: configures parser
bathyfacto/bathyfacto_config.py - Training methods → Masked ray data: configures manager
bathyfacto/bathyfacto_config.py - Training methods → BathyFacto model: configures model
bathyfacto/bathyfacto_config.py - Training methods → Train and evaluate: configures pipeline
bathyfacto/bathyfacto_config.py - Researcher → Nerfstudio runtime: starts training
README.md - Nerfstudio runtime → Train and evaluate: runs pipeline
- Train and evaluate → BathyFacto model: passes eval batch
bathyfacto/bathyfacto_pipeline.py - Masked ray data → BathyFacto model: provides masked rays
bathyfacto/bathyfacto_model.py - BathyFacto model → Radiance field: uses field
bathyfacto/bathyfacto_model.py - BathyFacto model → Two-media geometry: uses refraction geometry
bathyfacto/bathyfacto_model.py - Researcher → Point-cloud exporter: exports points
README.md - Point-cloud exporter → BathyFacto model: renders camera rays
bathyfacto/export.py - Point-cloud exporter → Masked ray data: attaches medium mask
bathyfacto/export.py - Point-cloud exporter → Frame normalization: uses frame transforms
bathyfacto/export.py - Point-cloud exporter → Bathymetric point cloud: writes PLY
bathyfacto/export.py