steps.ultralytics.model.evaluator

evaluator

Functions:

Name	Description
evaluate_ultralytics_model	Evaluate an Ultralytics model on a given dataset and log evaluation metrics.

evaluate_ultralytics_model(model, dataset)

Evaluate an Ultralytics model on a given dataset and log evaluation metrics.

This step handles evaluation for classification, object detection, and segmentation models trained with the Ultralytics framework. It:

• Retrieves the current training context from the pipeline.
• Chooses the appropriate predictor class based on the model's task type.
• Runs inference on the provided dataset in batches.
• Post-processes predictions into Picsellia-compatible format.
• Computes evaluation metrics and logs them to the experiment.

Supported tasks

• Classification
• Object Detection
• Segmentation

Parameters:

Name	Type	Description	Default
model	UltralyticsModel	The trained Ultralytics model to evaluate.	required
dataset	TBaseDataset	The dataset to evaluate the model on.	required

Returns:

Type	Description
None	None