shape¶

Active Shape Model (ASM) and Principal Component Analysis (PCA).

This module provides functionalities to learn the characteristic shapes of tracked objects from training data.

class ActiveShapeModel[source]¶

Bases: object

Constructs and manages an Active Shape Model in 2D.

Variables:

mean_shape (np.ndarray | None) – The mean vector of the aligned training shapes.
eigenvectors (np.ndarray | None) – The principal modes of shape variation.
eigenvalues (np.ndarray | None) – The variance explained by each principal component.

__init__()[source]¶

fit(training_data, ref_type='mean')[source]¶

Align training shapes and compute the shape model.

Parameters:

training_data (list[ndarray]) – List of N x 2 arrays representing contour vertices.
ref_type (str) – Alignment reference, either “mean” or “first”. Defaults to “mean”.

Return type:

None

Examples

>>> import numpy as np
>>> from super_slurpy.shape import ActiveShapeModel
>>> model = ActiveShapeModel()
>>> shapes = [np.random.rand(20, 2) for _ in range(5)]
>>> model.fit(training_data=shapes, ref_type="mean")

perform_pca(data_matrix)[source]¶

Perform Principal Component Analysis using SVD.

Parameters:: data_matrix (ndarray) – M x N matrix with M the vector length and N the datasets.
Returns:: Eigenvalues, Eigenvectors, and the mean vector.
Return type:: tuple[ndarray, ndarray, ndarray]

Examples

>>> import numpy as np
>>> from super_slurpy.shape import perform_pca
>>> data = np.random.rand(10, 5)
>>> evals, evecs, mean_vec = perform_pca(data_matrix=data)