xaitk_jatic.interop.object_detection.dataset

This module provides dataset classes for working with object detection data in the JATIC framework. It includes adapters for COCO-format datasets and general datasets for varying-sized images.

Classes:

JATICDetectionTarget: A dataclass for storing detection results, including bounding boxes, labels, and scores. COCOJATICObjectDetectionDataset: Converts a COCO dataset to be compatible with the JATIC object detection protocol. JATICObjectDetectionDataset: A generic wrapper for object detection datasets with varying-sized images.

Dependencies:

  • kwcoco: For working with COCO-format datasets.

  • numpy: For numerical operations.

  • maite.protocols: For object detection protocols.

  • PIL: For image processing.

Classes

COCOJATICObjectDetectionDataset

Dataset class to convert a COCO dataset to a dataset compliant with JATIC's Object Detection protocol.

COCOMetadata

TypedDict for COCO-detection datum-level metdata

JATICDetectionTarget

Dataclass for the datum-level JATIC output detection format.

JATICObjectDetectionDataset

Implementation of the JATIC Object Detection dataset wrapper for dataset images of varying sizes.