Project directory layout¶
Everything the extension writes is under <project>/celltune/:
celltune/
├── classifier-state.json # Multi-class model (features, classes, models, labels, normalisation)
├── composite-rules.json # Saved composite classification rules
├── marker-table.json # Channel mapping for review (§4.4)
├── binary-registry.json # List of binary classifiers and their state files
├── batch-shifts.json # Batch normalisation fit (§19)
├── labels_backup_YYYYMMDD_HHMMSS.json # Copy of the labels, saved before each Train
│
├── image-labels/ # Multi-class labels, one JSON per image
│ ├── slide1.json # { "<cellId>": "T-Cell", ... }
│ └── ...
│
├── binary-image-labels/<marker>/ # Same per-image JSON, one folder per binary marker
│ ├── CD3/slide1.json
│ └── ...
│
├── binary/ # Binary classifier state files
│ ├── CD3.json
│ ├── CD8.json
│ └── ...
│
├── image-sampled/ # Cell IDs already sampled for review
├── image-predictions/ # Multi-class predictions for each image (Review, Project Prediction Summary)
├── binary-image-predictions/ # Binary classifier predictions for each image, one folder per marker
├── logs/ # Training logs (last 20 kept)
All files are JSON. You can put celltune/ under version control (for example git) to share labels and review history.