Skip to content

PROJECT_STRUCTURE.md

Latest commit

 

History

History
78 lines (67 loc) · 2.69 KB

PROJECT_STRUCTURE.md

File metadata and controls

78 lines (67 loc) · 2.69 KB

Project Structure

The repository is organized to separate source code, model pipeline code, generated or binary artifacts, local data, runtime state, and documentation.

Canonical Layout

.
├── dermascan_app/
│   ├── app.py
│   ├── inference.py
│   ├── requirements.txt
│   └── templates/
├── ml_pipeline/
│   ├── config/
│   │   └── model_parameters.json
│   ├── evaluation/
│   │   └── evaluate_cnn.py
│   ├── registry/
│   │   └── progress_log.csv
│   └── training/
│       └── train_cnn.py
├── artifacts/
│   ├── models/
│   │   └── current/
│   │       └── melanoma_detector.keras
│   └── sample_inputs/
│       └── legacy_uploads/
├── data/
│   └── README.md
├── var/
│   ├── README.md
│   └── uploads/
├── docs/
│   ├── archived_previous_documents/
│   └── *.md
├── README.md
├── SECURITY.md
└── LICENSE

Naming Rationale

Name Rationale
dermascan_app/ Runtime application code only. This avoids the vague backend label and ties the folder to the product prototype.
ml_pipeline/ Training, evaluation, configuration, and registry files live together as the reproducibility surface.
ml_pipeline/config/ Parameters that define a model run belong in a predictable config home.
ml_pipeline/registry/ Historical metrics and model iteration records are treated as registry metadata.
artifacts/ Binary or non-source assets live outside app and pipeline source.
artifacts/models/current/ The app has a stable path for the currently packaged model.
artifacts/sample_inputs/legacy_uploads/ Old uploads are preserved as reviewable samples, not runtime state.
data/ Local datasets and manifests belong here, but datasets are intentionally not committed.
var/ Runtime state such as local uploads belongs here and should stay out of source control.
docs/archived_previous_documents/ Previous documentation is preserved without competing with the active docs.

Reproducibility Contract

Future model work should be reproducible from:

  • data/ dataset manifests
  • ml_pipeline/config/model_parameters.json
  • ml_pipeline/training/train_cnn.py
  • ml_pipeline/evaluation/evaluate_cnn.py
  • ml_pipeline/registry/progress_log.csv
  • artifacts/models/current/melanoma_detector.keras

Any new model release should record:

  • dataset version
  • train/validation/test split manifest
  • dependency versions
  • training command
  • evaluation command
  • threshold policy
  • generated metrics
  • model artifact path