steps.ultralytics.dataset.preparator

preparator

Functions:

Name	Description
prepare_ultralytics_dataset	Prepare and validate a dataset for training with the Ultralytics framework.

prepare_ultralytics_dataset()

Prepare and validate a dataset for training with the Ultralytics framework.

This step dynamically selects the appropriate dataset loading and formatting strategy based on the inference task type (classification, object detection, or segmentation) detected from the current experiment context.

Processing includes: - Loading COCO-style datasets for classification tasks and restructuring them into class-based folders. - Loading YOLO-style datasets for detection and segmentation tasks, followed by generating a data.yaml file. - Validating the dataset and optionally fixing annotation issues.

Returns:

Name	Type	Description
DatasetCollection	DatasetCollection	A dataset collection object ready for use in training pipelines.