🌟 API#

This part of the documentation covers all the interfaces of scivision.

scivision.io.reader.load_data_from_plugin(config: dict, branch: str = 'main') → Dataset#

Load a dataset from a data plugin (python package).

Parameters:

config (dict) – The loaded data.yml config.
branch (str, default = main) – Specify the name of a github branch if loading from github.

Returns:

The dataset to be visualised, loaded via xarray.

Return type:

xarray.Dataset

scivision.io.reader.load_dataset(path: PathLike, branch: str = 'main') → YAMLFileCatalog#

Load a dataset.

Parameters:

path (PathLike) – The filename, path or URL of an intake catalog, which links to a dataset.
branch (str, default = main) – Specify the name of a github branch if loading from github.

Returns:

An intake catalog object representing the loaded dataset (see intake.readthedocs).

Return type:

intake.catalog.local.YAMLFileCatalog

scivision.io.reader.load_pretrained_model(path: PathLike, branch: str = 'main', allow_install: bool = False, model_selection: str = 'default', load_multiple: bool = False, *args, **kwargs) → PretrainedModel#

Load a pre-trained model.

Parameters:

path (PathLike) – The filename, path or URL of a pretrained model description.
branch (str, default = main) – Specify the name of a github branch if loading from github.
allow_install (bool, default = False) – Allow installation of remote package via pip.
model_selection (str, default = default) – Specify the name of the model if there is > 1 in the model repo package.
load_multiple (bool, default = False) – Modifies the return to be a list of scivision.PretrainedModel’s.

Returns:

pretrained_model – The instantiated pre-trained model.

Return type:

scivision.PretrainedModel

class scivision.catalog.catalog.PandasCatalog(datasources=None, models=None, projects=None)#

compatible_datasources(model) → PandasQueryResult#

Return all datasources that are compatible with model

Parameters:: model (str or dict-like) – Any dictionary-like (including CatalogModelEntry) that has keys ‘format’, ‘tasks’ and ‘labels_required’, representing these properties of the model. If a string is passed, this is used to look up the model (in self._models).
Returns:: result – A QueryResult instance containing the datasources compatible with the given model (convertible to a dict or pd.DataFrame).
Return type:: QueryResult

compatible_models(datasource) → PandasQueryResult#

Return all models that are compatible with datasource

Parameters:: datasource (str or dict-like) – Any dictionary-like (including CatalogDatasourceEntry) that has keys ‘format’, ‘tasks’ and ‘labels_provided’, representing these properties of the datasource. If a string is passed, this is used to look up the datasource (in self._datasources).
Returns:: result – A QueryResult instance containing the models compatible with the given datasource (convertible to a dict or pd.DataFrame).
Return type:: QueryResult

validate()#

Extra validation

Raise ValueError if projects refer to nonexistent models or datasources

class scivision.vis.display.display_objects(image: ndarray, predictions: list, label_nums: bool = False)#

Display an image with colored bounding boxes for objects identified by a scivision object detection model.

Parameters:

image (np.ndarray) – A numpy ndarray representation of an image.
predictions (list) – A list of dictionaries, on for each detected object containing ‘score’, ‘label’ and ‘box’ keys. ‘score’ should be a float between 0 and 1. ‘label’ should be a string that labels the object. ‘box’ should be a dict with ‘xmin’, ‘ymin’, ‘xmax’ and ‘ymax’ keys, which denote the boundaries of each box (detected object).
label_nums (bool = False) – When True, bounding boxes are numbered in addition to being colored.

Returns:

The input image with colored bounding boxes and an accompanying pandas
dataframe legend showing corresponding labels and scores for each object.