Skip to content

Latest commit

 

History

History
72 lines (64 loc) · 3.24 KB

CHANGELOG.md

File metadata and controls

72 lines (64 loc) · 3.24 KB

Changelog

All notable changes to this project will be documented in this file.

[0.1.0] - 2026-01-31

Added

  • Initial implementation of TransUNet architecture for pancreas segmentation
  • Complete MONAI preprocessing pipeline with HU windowing
  • SlicingDataset for 2D slice extraction from 3D volumes
  • Hybrid loss function combining Dice Loss and Cross-Entropy Loss
  • Four Jupyter notebooks covering data exploration, model architecture, training, and evaluation
  • CLI interface in main.py for training and inference
  • Comprehensive documentation in README.md (NVIDIA personaplex-inspired design)
  • CLAUDE.md for AI assistant development guidance
  • Visualization utilities for CT slices and predictions
  • Professional README with badges, clear sections, and navigation
  • Assets folder structure for diagrams and visual documentation

Changed

  • README.md completely redesigned following NVIDIA personaplex style:
    • Added badges for Python, PyTorch, MONAI, License
    • Restructured with clear visual hierarchy
    • Added emoji icons for better navigation
    • Expanded Usage section with 4 notebook breakdowns
    • Added CLI interface documentation
    • Created detailed preprocessing pipeline table
    • Enhanced citation section with BibTeX
    • Professional contact & support section
  • Updated pyproject.toml description to be more descriptive
  • Fixed Python version requirement from >=3.13 to >=3.9 for better compatibility
  • Enhanced src/init.py with proper module exports and all
  • Corrected dataset size information to ~11.4GB (compressed) across all files
  • Fixed GitHub repository owner in README.md (ihatesea69)

Fixed

  • Corrected model.py line count in CLAUDE.md (18K → 533 lines)
  • Standardized dataset size documentation
  • Added known issues section to CLAUDE.md
  • Fixed inconsistent formatting in documentation

Project Structure

├── 📓 01-04_*.ipynb          # Educational notebooks
├── 📦 src/                   # Core implementation
│   ├── model.py              # TransUNet (533 lines)
│   ├── dataset.py            # SlicingDataset
│   ├── transforms.py         # MONAI pipeline
│   ├── loss.py               # HybridLoss
│   └── utils.py              # Visualization
├── 📁 assets/                # Documentation visuals
├── 💾 checkpoints/           # Model weights
├── 📊 outputs/               # Results
├── main.py                   # CLI entry point
├── pyproject.toml            # UV config
├── CHANGELOG.md              # Version history
└── README.md                 # Main documentation

Documentation Highlights

  • Clean, professional README inspired by NVIDIA's open-source projects
  • Comprehensive Usage guide with 3 access methods (Notebooks, CLI, Programmatic)
  • Visual architecture diagram placeholder in assets/
  • Detailed preprocessing pipeline table
  • Performance benchmarks section
  • Proper BibTeX citations for academic reference

Notes

  • Dataset must be downloaded via notebook 01 (~11.4GB)
  • Windows users should set num_workers=0 in DataLoader
  • Trained model checkpoints not included (empty checkpoints/ directory)
  • Architecture diagram in assets/ is currently ASCII placeholder - replace with PNG/SVG