Reference#

Inference#

predict(raw, mask, channel_names, mpp, *, ...)

Run the cell-type prediction pipeline.

PredictionResult(cell_types, probabilities, ...)

Structured predict() output when return_probabilities=True.

DCTConfig([zarr_path])

Inference-time configuration for deepcell_types.predict.

preprocess_fov(raw, mask, *, native_mpp, ...)

Canonical preprocessing for a single FOV.

PreprocessedFov(raw, mask, centroids, ...[, ...])

Result of preprocess_fov.

make_preprocessor(config)

Return a preprocess hook for deepcell_types.predict from a config.

apply_config(raw, channel_names, config)

Apply a bounded op pipeline to raw (C,H,W) -> (C,H,W) float32.

DEFAULT_CONFIG

Built-in mutable sequence.

Abstention#

ABSTENTION_LABEL

Sentinel cell-type name used for cells flagged as abstained.

Utilities#

download_model(*[, version])

Download the deepcell-types model checkpoint for local use.

download_baseline_checkpoint(name)

Download a baseline-model checkpoint (and any companion files).

download_training_data(*[, extract])

Download the public training-data corpus for deepcell-types (v1.1).

list_model_versions()

Return the available pre-trained model versions, default first.

list_baseline_names()

Return the available baseline identifiers, sorted.

list_supported_markers(*[, zarr_path])

Return the canonical marker names in the active registry, sorted.

list_supported_cell_types(*[, zarr_path])

Return the cell-type names the packaged registry recognizes, sorted.

resolve_supported_marker(marker, *[, zarr_path])

Resolve a marker name or alias to its canonical registry name.