steps.base.model.evaluator

evaluator

Functions:

Name	Description
evaluate_model	Perform evaluation of model predictions against ground truth annotations using Picsellia's ModelEvaluator.

evaluate_model(picsellia_predictions, inference_type, assets, output_dir)

Perform evaluation of model predictions against ground truth annotations using Picsellia's ModelEvaluator.

This step supports multiple inference types including classification, object detection, segmentation, and OCR. It compares the provided model predictions to the actual asset annotations, computes evaluation metrics, and stores the results in the specified output directory.

Parameters:

Name	Type	Description	Default
picsellia_predictions	List[Union[PicselliaClassificationPrediction, PicselliaRectanglePrediction, PicselliaPolygonPrediction, PicselliaOCRPrediction]]	List of model predictions corresponding to a supported inference type.	required
inference_type	InferenceType	The type of inference performed (e.g., CLASSIFICATION, DETECTION, SEGMENTATION, OCR).	required
assets	Union[List[Asset], MultiAsset]	The ground truth dataset assets to evaluate against.	required
output_dir	str	The path to the directory where evaluation results will be saved (e.g., confusion matrix, metrics report).	required