Helper Methods

Subpackages

• Transforms
  • Submodules
  • conex.helpers.transforms.encoders module
    • SimplePoisson
      • SimplePoisson.training
    • Poisson
      • Poisson.training
    • Intensity2Latency
      • Intensity2Latency.training
  • conex.helpers.transforms.masks module
    • GridEraseMask
    • GridKeepMask
    • GridCropMask
  • conex.helpers.transforms.misc module
    • UnsqueezeTransform
      • UnsqueezeTransform.training
    • SqueezeTransform
      • SqueezeTransform.training
    • SwapTransform
      • SwapTransform.training
    • DeviceTransform
      • DeviceTransform.training
    • Conv2dFilter
      • Conv2dFilter.training
    • AbsoluteTransform
      • AbsoluteTransform.training
    • DivideSignPolarity
      • DivideSignPolarity.training

Submodules

conex.helpers.data module

class conex.helpers.data.LocationDataset(dataset, pre_transform=None, post_transform=None, location_transform=None, target_transform=None)

Bases: Dataset

A custom dataset class for data with triple image, location, label nature.

Parameters:

• dataset (Dataset) – An Instance of a dataset

• pre_transform (Transform) – A transformation to apply on images. If given, transformation should return a (image, location) tuple.

• post_transform (Transform) – A Transformation that applies on images. Suitable for encodings.

• location_transform (Transform) – A Transformation applies on location data. Suitable for encodings.

• target_transform (Transform) – A Transformation applies on labels.

conex.helpers.filters module

conex.helpers.filters.DoGFilter(size, sigma_1, sigma_2, step=1.0, zero_mean=False, one_sum=False, device=None, dtype=None)

Difference of Gaussians. Makes a square mono-colored DoG filter.

Parameters:

• size (int) – Filter size.

• sigma_1 (float) – First standard deviation.

• sigma_2 (float) – Second standard deviation.

• step (float, optional) – Scaling factor for axes. Defaults to 1.0.

• zero_mean (bool, optional) – Whether to scale negative values in order to have zero mean. Defaults to False.

• one_sum (bool, optional) – Whether to divide values in order to have maximum possible dot product equal to one. Defaults to False.

• device (str, optional) – Device to locate filter on. Defaults to None.

• dtype (dtype, optional) – Datatype of desired filter. Defaults to None.

Returns:

the desired DoG filter

Return type:

tensor

conex.helpers.filters.GaborFilter(size, labda, theta, sigma, gamma, step=1.0, zero_mean=False, one_sum=False, device=None, dtype=None)

Gabor filter Makes a square mono-colored Gabor filter.

Parameters:

• size (int) – Filter size.

• labda (float) – The wavelength of the filter.

• theta (float) – The orientation of the filter.

• sigma (float) – The standard deviation of the filter.

• gamma (float) – The aspect ratio for the filter.

• step (float, optional) – Scaling factor for axes. Defaults to 1.0.

• zero_mean (bool, optional) – Whether to scale negative values in order to have zero mean. Defaults to False.

• one_sum (bool, optional) – Whether to divide values in order to have maximum possible dot product equal to one. Defaults to False.

• device (str, optional) – Device to locate filter on. Defaults to None.

• dtype (dtype, optional) – Datatype of desired filter. Defaults to None.

Returns:

the desired Gabor filter

Return type:

tensor