Load new datasets

mira.datasets.load_coco(annotations_file, image_dir, categories=None)[source]

Obtain a scene collection from a COCO JSON file.

Parameters

  • annotations_file (str) – The annotation file to load

  • image_dir (str) – The directory in which to look for images

  • categories (Optional[Categories]) – The annotation configuration to use. If None, it is inferred from the annotations category file.

Return type

SceneCollection

Returns

A scene collection

mira.datasets.load_voc(filepaths, categories, image_dir=None)[source]

Read a scene from a VOC XML annotation file. Remaining arguments passed to scene constructor.

Parameters

  • filepaths (List[str]) – A list of VOC files to read

  • image_folder – Folder in which to look for images. Defaults to same folder as XML file prepended to the folder specified in the XML file.

  • categories (Categories) – The annotation configuration to use.

Return type

SceneCollection

Returns

A new scene collection, one scene per VOC file