How ahobe/rockoders_sem_image_assessment works
AssesSEM is a command-line workflow for segmenting minerals and pore spaces in SEM thin-section images and reporting their relative areas. Users choose datasets, model, predictor, and image count; the pipeline pairs BSE/CL images (optionally mineral-map images), predicts per-pixel classes, saves segmentation images, and writes CSV summaries. The graph includes the documented model and modality options; unsampled model files are represented by their listed directory, and edges are limited to visible imports/calls or README statements.
Main entry points: pipelines.py Prediction pipeline · IO.py Image and file I/O · model_manipulation.py Model loading · use_cases.py Image prediction use case
Components 17
Workflow and I/O
- CLI runner
Data/run.py - Prediction pipeline
src/assesSEM/pipelines.py - Input selection
src/assesSEM/get_user_input.py - Image and file I/O
src/assesSEM/IO.py - SEM image datasets
Data
Model and prediction
- Model loading
src/assesSEM/model_manipulation.py - U-Net and inputs
src/assesSEM/unet.py - Trained weights
src/assesSEM/models - Image prediction use case
src/assesSEM/use_cases.py - Prediction strategies
src/assesSEM/predictors.py - Smooth tiled predictor
src/assesSEM/smooth_tiled_predictions.py - Image slicing
src/assesSEM/predicting/slicing.py
Segmentation results
- Class and area processing
src/assesSEM/postprocessing.py - Segmentation rendering
src/assesSEM/plotting.py - Segmentation outputs
Data/dataset1/CL_segmented - Area summary CSV
Data/dataset1/CL_segmented
Other components
- Dataset user
Connections 19
- Dataset user → CLI runner: starts
README.md - CLI runner → Prediction pipeline: runs
Data/run.py - Prediction pipeline → Input selection: requests choices
src/assesSEM/pipelines.py - Prediction pipeline → Image and file I/O: prepares and saves
src/assesSEM/pipelines.py - Prediction pipeline → SEM image datasets: selects image files
src/assesSEM/pipelines.py - Prediction pipeline → Model loading: loads model
src/assesSEM/pipelines.py - Model loading → U-Net and inputs: builds model
src/assesSEM/model_manipulation.py - Model loading → Trained weights: loads weights
README.md - Prediction pipeline → Image prediction use case: requests predictions
src/assesSEM/pipelines.py - Image prediction use case → U-Net and inputs: builds model input
src/assesSEM/use_cases.py - Image prediction use case → Prediction strategies: dispatches prediction
src/assesSEM/use_cases.py - Prediction strategies → Smooth tiled predictor: uses tiled prediction
src/assesSEM/predictors.py - Prediction strategies → Image slicing: gets slice boxes
src/assesSEM/predictors.py - Prediction pipeline → Class and area processing: converts and measures
src/assesSEM/pipelines.py - Prediction pipeline → Segmentation outputs: saves predictions
src/assesSEM/pipelines.py - Prediction pipeline → Area summary CSV: writes area summary
src/assesSEM/pipelines.py - Image and file I/O → Input selection: checks output folder
src/assesSEM/IO.py - Image and file I/O → Segmentation rendering: gets color map
src/assesSEM/IO.py - U-Net and inputs → Image and file I/O: reads and normalizes
src/assesSEM/unet.py