Datasets Module#

Datasets module.

landmarker.datasets.get_cepha_heatmap_datasets(path_dir, train_transform=None, inference_transform=None, sigma=1, junior=False, single_dataset=False, cv=False, **kwargs)[source]#

Returns a HeatmapDataset with the ISBI 2015 cephalogram challenge dataset. The dataset is split into train, test1 and test2. The same approach as in “CephaNN: A Multi-Head Attention Network for Cephalometric Landmark Detection” - JIAHOONG QIAN et al. is used.

Parameters:

  • path_dir (str) – The path to the directory where the dataset should be stored.

  • train_transform (Optional[Callable], optional) – A transformation to apply to the images and landmarks during training. Defaults to None.

  • inference_transform (Optional[Callable], optional) – A transformation to apply to the images and landmarks during inference. Defaults to None.

  • sigma (int, optional) – The sigma value for the gaussian kernel. Defaults to 1.

  • junior (bool, optional) – Whether to use the junior or senior annotator. Defaults to False.

  • single_dataset (bool, optional) – Whether to return a single dataset with all images and landmarks. Defaults to False.

  • **kwargs – Additional keyword arguments for the HeatmapDataset.

  • cv (bool) –

Return type:

HeatmapDataset | tuple[HeatmapDataset, HeatmapDataset, HeatmapDataset] | tuple[HeatmapDataset, HeatmapDataset, HeatmapDataset, HeatmapDataset]

landmarker.datasets.get_cepha_landmark_datasets(path_dir, train_transform=None, inference_transform=None, store_imgs=True, dim_img=None, junior=False, single_dataset=False, cv=False)[source]#

Returns a LandmarkDataset objects with the CEPH dataset, a combination of the ISBI 2014 & 2015 challenges. The dataset is split into train, test1 and test2. The same approach as in “CephaNN: A Multi-Head Attention Network for Cephalometric Landmark Detection” - JIAHOONG QIAN

et al. is used.

Parameters:

path_dir (str) –

Return type:

LandmarkDataset | tuple[LandmarkDataset, LandmarkDataset, LandmarkDataset] | tuple[LandmarkDataset, LandmarkDataset, LandmarkDataset, LandmarkDataset]

landmarker.datasets.get_cepha_mask_datasets(path_dir, train_transform=None, inference_transform=None, junior=False, single_dataset=False, cv=False, **kwargs)[source]#

Returns a MaskDataset with the ISBI 2015 cephalogram challenge dataset. The dataset is split into train, test1 and test2. The same approach as in “CephaNN: A Multi-Head Attention Network for Cephalometric Landmark Detection” - JIAHOONG QIAN et al. is used.

Parameters:

  • path_dir (str) – The path to the directory where the dataset should be stored.

  • train_transform (Optional[Callable], optional) – A transformation to apply to the images and masks during training. Defaults to None.

  • inference_transform (Optional[Callable], optional) – A transformation to apply to the images and masks during inference. Defaults to None.

  • junior (bool, optional) – Whether to use the junior or senior annotator. Defaults to False.

  • single_dataset (bool, optional) – Whether to return a single dataset with all images and landmarks. Defaults to False.

  • **kwargs – Additional keyword arguments for the MaskDataset.

  • cv (bool) –

Return type:

MaskDataset | tuple[MaskDataset, MaskDataset, MaskDataset] | tuple[MaskDataset, MaskDataset, MaskDataset, MaskDataset]

landmarker.datasets.get_cepha_patch_datasets(path_dir, index_landmark=0, train_transform=None, inference_transform=None, store_imgs=True, junior=False, single_dataset=False, cv=False, **kwargs)[source]#

Returns a PatchDataset objects with the CEPH dataset, a combination of the ISBI 2014 & 2015 challenges. The dataset is split into train, test1 and test2. The same approach as in “CephaNN: A Multi-Head Attention Network for Cephalometric Landmark Detection” - JIAHOONG QIAN

et al. is used.

Parameters:

  • path_dir (str) –

  • index_landmark (int) –