Query SDK

Query the Datapoints / Predictions / Groundtruths

To query the Datapoints, Predictions or Groundtruths, you can use the select_datapoints select_predictions , select_groundtruths utilities. The output is a pandas DataFrame

from dioptra.lake.utils import (
    select_datapoints,
    select_predictions,
    select_groundtruths
)

def select_datapoints( # similarly select_predictions and select_groundtruths
    fields: List[str] -> ['*'],
    filters: List[object],
    limit: Optional[int],
    order_by: Optional[str],
    desc: Optional[bool]
)

Parameter	Description
fields	the list of fields to be retrieved. To query sub fields, use a jq format. For example: image_metadata.uri
filters	A list of filters to be used to query the data. Should follow the format { 'left': 'my_field', 'op': '=', 'right: 'my_value' } Supported operators are =, !=, >, <, >=, <=, in, not in, like, not like
limit	limit to the number of datapoints to query
order_by	the name of a field to order by when doing a limit This is to control the order on which the limit is applied
desc	whether the order should be desc or asc

Joining query results

To join query results in a single DataFrame, you can use the join_on_datapoints utility

from dioptra.lake.utils import join_on_datapoints

def join_on_datapoints(
    datapoints: DataFrame,
    predictions: DataFrame,
    groundtruths: DataFrame,
)

Parameter	Description
datapoints	a DataFrame containing the datapoints
predictions	a DataFrame containing the predictions
groundtruths	a DataFrame containing the groundtruths

Deleting

To delete datapoints, predictions or ground truths, use the delete_xxx functions. Note that deleting a datapoint will delete all attachment: groundtruths, tags, predictions.

from dioptra.lake.utils import delete_datapoints, delete_predictions, delete_groundtruths

def delete_datapoints(filters, limit=None, order_by=None, desc=None)

def delete_predictions(prediction_ids)

def delete_groundtruths(groundtruth_ids)