Napari Entry
Labelprop Index / Labelprop / Napari Entry
Auto-generated documentation for labelprop.napari_entry module.
Attributes
-
package_dir- Get package directory: pathlib.Path(file).parent.absolute() -
home- Get user home directory: pathlib.Path.home() - get_ckpt_dir
- get_fields
- pretrain
- propagate_from_ckpt
- propagate_from_fields
- resample
- train_and_infer
- train_dataset
get_ckpt_dir
Show source in napari_entry.py:24
Signature
def get_ckpt_dir(): ...
get_fields
Show source in napari_entry.py:227
Get successive fields from the LabelProp model.
Arguments
imgTensor - The input image.ckptstr - The path to the checkpoint file.maskTensor, optional - The mask image. Defaults to None.z_axisint, optional - The axis representing the z-dimension. Defaults to 2.selected_sliceslist, optional - The list of selected slices. Defaults to None.
Returns
Tensor- The concatenated fields_up.Tensor- The concatenated fields_down.Tensor- The input image (X).Tensor- The target image (Y).
Signature
def get_fields(img, ckpt, mask=None, z_axis=2, selected_slices=None): ...
pretrain
Show source in napari_entry.py:153
Pretrains a model using unsupervised learning on a list of images.
Arguments
img_listlist - List of image paths.shapeint - Desired shape of the images after resizing.z_axisint, optional - Axis along which to slice the images. Defaults to 2.output_dirstr, optional - Directory to save the trained model checkpoint. Defaults to '~/label_prop_checkpoints'.namestr, optional - Name of the trained model checkpoint. Defaults to an empty string.max_epochsint, optional - Maximum number of training epochs. Defaults to 100.
Signature
def pretrain(
img_list,
shape,
z_axis=2,
output_dir="~/label_prop_checkpoints",
name="",
max_epochs=100,
): ...
propagate_from_ckpt
Show source in napari_entry.py:42
Propagates labels from a given checkpoint using the LabelPropDataModule and inference.
Arguments
img (ndarray or str): Input image data or path to the image file.
mask (ndarray or str): Input mask data or path to the mask file.
- checkpoint str - Path to the checkpoint file.
- shape int, optional - Shape of the input image and mask. Defaults to 304.
- z_axis int, optional - Axis along which the slices are selected. Defaults to 2.
- label str, optional - Label to propagate. Defaults to 'all'.
hints (ndarray or str, optional): Input hints data or path to the hints file. Defaults to None.
- **kwargs - Additional keyword arguments to be passed to the inference function.
Returns
tuple- A tuple containing the propagated labels for the up direction, down direction, and fused direction.
Signature
def propagate_from_ckpt(
img, mask, checkpoint, shape=304, z_axis=2, label="all", hints=None, **kwargs
): ...
propagate_from_fields
Show source in napari_entry.py:179
Propagates labels from given fields to the input image.
Arguments
img (str or ndarray): Path to the input image or the image array itself.
mask (str or ndarray): Path to the mask image or the mask array itself.
- fields_up ndarray - Array of fields for upward propagation.
- fields_down ndarray - Array of fields for downward propagation.
- shape int - Desired shape of the propagated labels.
- z_axis int, optional - Axis along which the slices are selected. Defaults to 2.
- selected_slices list, optional - List of selected slices. Defaults to None.
- kwargs dict, optional - Additional keyword arguments for propagation.
Returns
tuple or tuple of ndarrays: Tuple containing the propagated labels for upward propagation, downward propagation, and fused propagation. If 'return_weights' is True in kwargs, the tuple also contains the weights used for propagation.
Signature
def propagate_from_fields(
img, mask, fields_up, fields_down, shape, z_axis=2, selected_slices=None, kwargs={}
): ...
resample
Show source in napari_entry.py:32
Resample a label tensor to the given size
Signature
def resample(Y, size): ...
train_and_infer
Show source in napari_entry.py:88
Trains a model and performs inference on the given image and mask.
Arguments
img (ndarray or str): The input image data or path to the image file.
mask (ndarray or str): The input mask data or path to the mask file.
- pretrained_ckpt str - The path to the pretrained checkpoint file.
- shape int - The desired shape of the input image and mask.
- max_epochs int - The maximum number of training epochs.
- z_axis int, optional - The axis along which the slices are selected. Defaults to 2.
- output_dir str, optional - The directory to save the output checkpoint file. Defaults to '~/label_prop_checkpoints'.
- name str, optional - The name of the output checkpoint file. Defaults to an empty string.
- pretraining bool, optional - Whether to perform pretraining. Defaults to False.
hints (ndarray or str, optional): The input hints data or path to the hints file. Defaults to None.
- **kwargs - Additional keyword arguments for training and inference.
Returns
tuple- A tuple containing the upsampled, downsampled, and fused predictions as numpy arrays.
Signature
def train_and_infer(
img,
mask,
pretrained_ckpt,
shape,
max_epochs,
z_axis=2,
output_dir="~/label_prop_checkpoints",
name="",
pretraining=False,
hints=None,
**kwargs
): ...
train_dataset
Show source in napari_entry.py:259
Trains a dataset using label propagation.
Arguments
img_listlist - List of image file paths.mask_listlist - List of mask file paths.pretrained_ckptstr - Path to the pretrained checkpoint.shapeint - Shape of the input images.max_epochsint - Maximum number of training epochs.z_axisint, optional - Z-axis index. Defaults to 2.output_dirstr, optional - Output directory for saving checkpoints. Defaults to '~/label_prop_checkpoints'.namestr, optional - Name of the saved checkpoint. Defaults to ''.**kwargs- Additional keyword arguments.
Returns
str- Path to the best checkpoint.
Signature
def train_dataset(
img_list,
mask_list,
pretrained_ckpt,
shape,
max_epochs,
z_axis=2,
output_dir="~/label_prop_checkpoints",
name="",
**kwargs
): ...