How styx4d/galet works
GALET is a QGIS client and localhost Mask R-CNN server for segmenting gravel grains in ordinary images and georeferenced orthomosaics. The principal workflow is to submit image data from a QGIS processing tool, run instance segmentation on the server, then use the returned masks and boxes to produce GIS results. Orthomosaic processing adds spatial clipping, tiling, and duplicate merging; image processing has its own raster workflow. The repository also includes a grain-specific training setup. The processing excerpts are partial, so uncertain links between later geospatial steps are omitted.
Main entry points: Qgis_processing_ORTHO.py Ortho tool · Qgis_processing_IMAGE.py Image tool · Galet_Server.py TCP request handler · ShareArrays.py Array transfer
Components 20
QGIS processing
- Image tool
Qgis_processing_IMAGE.py - Image tiling
Qgis_processing_IMAGE.py - Image GIS output
Qgis_processing_IMAGE.py - Ortho tool
Qgis_processing_ORTHO.py - Ortho clipping
Qgis_processing_ORTHO.py - Ortho tiling
Qgis_processing_ORTHO.py - Duplicate merging
Qgis_processing_ORTHO.py - Ortho GIS output
Qgis_processing_ORTHO.py
Inference service
- TCP request handler
Galet_Server.py - Model setup
Galet_Server.py
Mask R-CNN
- Segmentation model
GALET_RCNN_V3/mrcnn/model.py - Network settings
GALET_RCNN_V3/mrcnn/config.py - Detection utilities
GALET_RCNN_V3/mrcnn/utils.py
Training support
- Grain configuration
GALET_RCNN_V3/grain.py - Grain dataset
GALET_RCNN_V3/grain.py - Training runtime
GALET_RCNN_V3/mrcnn/model.py
Shared-memory transport
- Array transfer
ShareArrays.py
Other components
- GIS user
- Local TCP
- Model weights
Connections 28
- GIS user → Image tool: runs
- GIS user → Ortho tool: runs
- Image tool → Image tiling: prepares tiles
- Image tiling → Array transfer: shares image
- Image tool → TCP request handler: requests inference
- Image tool → Array transfer: reads results
- Image tool → Image GIS output: writes GIS output
- Ortho tool → Ortho clipping: clips raster
- Ortho tool → Ortho tiling: builds grid
- Ortho tiling → Array transfer: shares tile data
- Ortho tool → TCP request handler: requests inference
- Ortho tool → Array transfer: reads results
- Ortho tool → Duplicate merging: merges detections
- Ortho tool → Ortho GIS output: writes GIS output
- TCP request handler → Local TCP: serves requests
- Image tool → Local TCP: sends request
- Ortho tool → Local TCP: sends request
- TCP request handler → Array transfer: reads image
- TCP request handler → Array transfer: shares detections
- TCP request handler → Segmentation model: runs detect
- Model setup → Segmentation model: creates model
- Model setup → Grain configuration: uses settings
- Model setup → Model weights: loads
- Segmentation model → Network settings: uses settings
- Segmentation model → Detection utilities: uses utilities
- Grain dataset → Detection utilities: extends dataset base
- Grain dataset → Training runtime: provides training data
- Grain configuration → Training runtime: configures training