mmv_im2im.postprocessing package

Submodules

mmv_im2im.postprocessing.basic_collection module

mmv_im2im.postprocessing.basic_collection.extract_segmentation(im: ndarray | Tensor, channel: int, cutoff: float | str | None = None, batch_dim: bool = True) → ndarray

extract segmentation from a prediction

Parameters:

im: ndarray or torch.Tensor

the multi-class prediction (1, C, W, H) or (1, C, Z, Y, X)

channel: int

which channel to select

cutoff: float or str

either a fixed cutoff value or a segmentation method from skimage, default is None (do not apply any cutoff)

batch_dim: bool

whether there is a batch dimension (default is True)

mmv_im2im.postprocessing.basic_collection.generate_classmap(im: ndarray | Tensor) → ndarray

generate the segmentation classmap from model prediction

Parameters:

im: ndarray or torch.Tensor

the multi-class prediction (1, C, W, H) or (1, C, Z, Y, X)

mmv_im2im.postprocessing.embedseg_cluster module

class mmv_im2im.postprocessing.embedseg_cluster.Cluster_2d(grid_y, grid_x, pixel_y, pixel_x)

Bases: object

cluster(prediction, n_sigma=2, seed_thresh=0.5, min_mask_sum=5, min_unclustered_sum=0, min_object_size=5)

cluster_with_gt(prediction, instance, n_sigma=1)

class mmv_im2im.postprocessing.embedseg_cluster.Cluster_3d(grid_z, grid_y, grid_x, pixel_z, pixel_y, pixel_x, one_hot=False)

Bases: object

cluster(prediction, n_sigma=3, seed_thresh=0.5, min_mask_sum=64, min_unclustered_sum=0, min_object_size=3)

cluster_with_gt(prediction, instance, n_sigma=1)

mmv_im2im.postprocessing.embedseg_cluster.degrid(meter, grid_size, pixel_size)

mmv_im2im.postprocessing.embedseg_cluster.generate_instance_clusters(pred: ndarray | Tensor, grid_x: int = 768, grid_y: int = 768, pixel_x: int = 1, pixel_y: int = 1, n_sigma: int = 2, seed_thresh: float = 0.5, min_mask_sum: int = 10, min_unclustered_sum: int = 10, min_object_size: int = 10, grid_z: int = 32, pixel_z: int = 1)

Module contents