How anjanisu/dfutissuesegnet works
This repository trains four-class DFU tissue segmentation models using labeled images and progressively pseudo-labeled unlabeled images. The intended workflow is supervised training, Phase 1 pseudo-labeling and semi-supervised refinement, then Phase 2 pseudo-labeling and refinement; evaluation and loss-curve generation provide comparison and diagnostics. The README documents the model architecture and full workflow. Source excerpts show training, inference, and evaluation behavior, but the master script references a Codes/ directory absent from the tree, so its runnable orchestration is uncertain.
Main entry points: semisupervised_training_phase1_local.py Phase 1 semi-supervised · run_master_pipeline.py Pipeline launcher · semisupervised_training_phase2_local.py Phase 2 semi-supervised · pseudo_label_generation_phase2_local.py Phase 2 pseudo-labels
Components 17
Model training
- Supervised training
codx/supervised_training_local.py - Segmentation model
- Phase 1 semi-supervised
codx/semisupervised_training_phase1_local.py - Phase 2 semi-supervised
codx/semisupervised_training_phase2_local.py - Pipeline launcher
codx/run_master_pipeline.py
Pseudo-labeling
- Phase 1 pseudo-labels
codx/pseudo_label_generation_phase1_local.py - Phase 2 pseudo-labels
codx/pseudo_label_generation_phase2_local.py
Models and results
- Model checkpoints
- Pseudo-label files
- Training metrics
- Phase results log
output.txt
Evaluation and diagnostics
- Phase evaluation
codx/evaluate_all_phases.py - Combined loss curves
codx/generate_combined_loss_curves.py - Per-run loss curves
codx/generate_loss_curves_local.py - Metric search
codx/find_85_90_metrics_local.py
Other components
- Researcher
- DFUTissue data
Connections 30
- Researcher → Pipeline launcher: starts
README.md - DFUTissue data → Supervised training: labeled images
README.md - Supervised training → Segmentation model: trains
README.md - Supervised training → Model checkpoints: saves checkpoint
codx/supervised_training_local.py - Supervised training → Training metrics: writes metrics
codx/supervised_training_local.py - Model checkpoints → Phase 1 pseudo-labels: loads model
codx/pseudo_label_generation_phase1_local.py - DFUTissue data → Phase 1 pseudo-labels: unlabeled images
codx/pseudo_label_generation_phase1_local.py - Phase 1 pseudo-labels → Pseudo-label files: writes labels
codx/pseudo_label_generation_phase1_local.py - Pseudo-label files → Phase 1 semi-supervised: pseudo-labeled data
README.md - DFUTissue data → Phase 1 semi-supervised: labeled data
codx/semisupervised_training_phase1_local.py - Phase 1 semi-supervised → Segmentation model: trains
README.md - Phase 1 semi-supervised → Model checkpoints: saves checkpoint
codx/semisupervised_training_phase1_local.py - Phase 1 semi-supervised → Training metrics: writes metrics
README.md - Model checkpoints → Phase 2 pseudo-labels: loads model
codx/pseudo_label_generation_phase2_local.py - Pseudo-label files → Phase 2 pseudo-labels: excludes Phase 1 labels
codx/pseudo_label_generation_phase2_local.py - DFUTissue data → Phase 2 pseudo-labels: remaining images
codx/pseudo_label_generation_phase2_local.py - Phase 2 pseudo-labels → Pseudo-label files: writes labels
codx/pseudo_label_generation_phase2_local.py - Pseudo-label files → Phase 2 semi-supervised: pseudo-labeled data
README.md - Phase 2 semi-supervised → Segmentation model: trains
README.md - Phase 2 semi-supervised → Model checkpoints: saves checkpoint
codx/semisupervised_training_phase2_local.py - Phase 2 semi-supervised → Phase results log: appends results
codx/semisupervised_training_phase2_local.py - DFUTissue data → Phase evaluation: test set
codx/evaluate_all_phases.py - Model checkpoints → Phase evaluation: evaluates models
codx/evaluate_all_phases.py - Training metrics → Combined loss curves: plots histories
codx/generate_combined_loss_curves.py - Training metrics → Per-run loss curves: plots histories
codx/generate_loss_curves_local.py - Training metrics → Metric search: searches values
codx/find_85_90_metrics_local.py - Pipeline launcher → Phase 1 semi-supervised: invokes phase
codx/run_master_pipeline.py - Pipeline launcher → Phase 2 pseudo-labels: invokes phase
codx/run_master_pipeline.py - Pipeline launcher → Phase 2 semi-supervised: invokes phase
codx/run_master_pipeline.py - Pipeline launcher → Combined loss curves: invokes plotting
codx/run_master_pipeline.py