Skip to content

steps.base.model.prediction_converter

prediction_converter

Functions:

Name Description
convert_predictions_to_coco

Convert a list of Picsellia predictions into COCO format annotations.

Attributes:

Name Type Description
PredictionType

PredictionType = Union[PicselliaClassificationPrediction, PicselliaRectanglePrediction, PicselliaPolygonPrediction] module-attribute

convert_predictions_to_coco(predictions, dataset, use_id=False)

Convert a list of Picsellia predictions into COCO format annotations.

Supports: - Classification (as single-class boxes covering full image) - Object detection (rectangle) - Segmentation (polygon)

Parameters:

Name Type Description Default

predictions

list[PredictionType]

List of predictions (classification, detection or segmentation)

required

dataset

CocoDataset

Dataset containing image + category info

required

use_id

bool

If True, match images using asset.id_with_extension instead of asset.filename

False

Returns:

Type Description
CocoDataset

Updated CocoDataset