How msantacesaria/ktc2023_pnpe2e works
This repository implements an EIT challenge reconstruction application: it uses measured voltage data to produce pixel-grid conductivity segmentations, combining a learned end-to-end model with a plug-and-play Gauss–Newton reconstruction and mask-based postprocessing. The README also documents graph-U-Net denoising and synthetic inclusion generation. Edges reflect the shown main/helper calls and README-described workflow; unsampled details are left unconnected where their wiring cannot be verified.
Main entry points: main.py Reconstruction entry point · helper_fncs.py End-to-end model · KTCMeshing.py Finite-element mesh · KTCFwd.py EIT forward solver
Components 20
Reconstruction workflow
- Reconstruction entry point
main.py - Segmented reconstruction
- PnP Gauss–Newton
main.py
Learned reconstruction
- End-to-end model
helper_fncs.py - Learned linear inverse
helper_fncs.py - Mesh interpolator
helper_fncs.py - Image U-Net
main.py - Graph U-Net denoiser
main.py
EIT numerical core
- EIT forward solver
KTCFwd.py - Finite-element mesh
KTCMeshing.py - Premade mesh data
Mesh_sparse.mat - Circular mesh routines
KTCcircmesh.py - Smoothness prior
KTCRegularization.py
Segmentation and data
- Difficulty masks
helper_fncs.py - Threshold segmentation
KTCScoring.py - Mesh interpolation helper
KTCAux.py - Synthetic inclusion generator
KTCAux_luca.py - Reconstruction plotting
KTCPlotting.py
Other components
- Challenge user
- Voltage measurements
Connections 21
- Challenge user → Reconstruction entry point: runs
README.md - Challenge user → Voltage measurements: provides
README.md - Reconstruction entry point → Voltage measurements: loads
main.py - Reconstruction entry point → Difficulty masks: gets masks
main.py - Reconstruction entry point → End-to-end model: runs
main.py - End-to-end model → Learned linear inverse: applies
helper_fncs.py - End-to-end model → Mesh interpolator: interpolates
helper_fncs.py - End-to-end model → Image U-Net: applies
helper_fncs.py - Reconstruction entry point → PnP Gauss–Newton: runs
main.py - PnP Gauss–Newton → Graph U-Net denoiser: denoises with
README.md - PnP Gauss–Newton → EIT forward solver: uses forward model
README.md - Reconstruction entry point → EIT forward solver: configures solver
main.py - Reconstruction entry point → Smoothness prior: builds prior
main.py - Reconstruction entry point → Finite-element mesh: loads mesh
main.py - Finite-element mesh → Circular mesh routines: imports
KTCMeshing.py - Reconstruction entry point → Premade mesh data: loads
main.py - Reconstruction entry point → Threshold segmentation: segments reconstruction
README.md - Reconstruction entry point → Segmented reconstruction: writes
README.md - Reconstruction entry point → Reconstruction plotting: imports
main.py - Reconstruction entry point → Mesh interpolation helper: imports
main.py - Synthetic inclusion generator → Finite-element mesh: generates shapes on