Module src.app.Clustering.ClusterVoronoiTesselation

Classes

class ClusterVoronoiTesselation (prev_module, num_iterations=15)

Applies Voronoi Tessellation to image clusters

This class uses the clustering result and subdivides it via Voronoi Tessellation. In addition it runs CVT to evenly distribute the cells in the given space.

Attributes

_num_iterations

Number of iterations to use for CVT (int)

Source code

class ClusterVoronoiTesselation(Module):
    """Applies Voronoi Tessellation to image clusters

    This class uses the clustering result and subdivides it via Voronoi Tessellation.
    In addition it runs CVT to evenly distribute the cells in the given space.

    Attributes:
        _num_iterations: Number of iterations to use for CVT (int)
    """
    def __init__(self, prev_module, num_iterations=15):
        super().__init__('ClusterVoronoiTesslation', prev_module)
        self._num_iterations = num_iterations

    def run(self):
        super().run()
        points = np.concatenate((np.array([self._data['cx']]).T, np.array([self._data['cy']]).T), axis=1)

        centroids, voronoi = clv.cvt(points, [0, 1, 0, 1], self._num_iterations)
        self.create_result(centroids, voronoi)

    def create_result(self, centroids, voronoi):
        """Arranges tessellation results in a new result object.

        Args:
            centroids: List of centroids of the voronoi cells (list of lists)
            voronoi: Tessellation result (object)
        """
        self._result = {
            'clusters': [],
            'voronoi': voronoi,
            'centroids': centroids
        }

        num_unique_labels = len(np.unique(self._data['labels']))
        points = np.concatenate((np.array([self._data['x']]).T, np.array([self._data['y']]).T), axis=1)
        centers = np.concatenate((np.array([self._data['cx']]).T, np.array([self._data['cy']]).T), axis=1)
        for label in range(num_unique_labels):
            p = [p for i, p in enumerate(points) if self._data['labels'][i] == label]
            imgs = [img for i, img in enumerate(self._data['images']) if self._data['labels'][i] == label]
            self._result['clusters'].append({
                'images': imgs,
                'label': label,
                'points': p,
                'center': centers[label],
            })

    def visualize(self):
        result = self.get_module_results()
        plt.figure()
        voronoi_plot_2d(result['voronoi'])
        plt.xlim((0, 1))
        plt.ylim((0, 1))
        plt.show()

Ancestors

• Module

Methods

def create_result(self, centroids, voronoi)

Arranges tessellation results in a new result object.

Args

centroids

List of centroids of the voronoi cells (list of lists)

voronoi

Tessellation result (object)

Source code

def create_result(self, centroids, voronoi):
    """Arranges tessellation results in a new result object.

    Args:
        centroids: List of centroids of the voronoi cells (list of lists)
        voronoi: Tessellation result (object)
    """
    self._result = {
        'clusters': [],
        'voronoi': voronoi,
        'centroids': centroids
    }

    num_unique_labels = len(np.unique(self._data['labels']))
    points = np.concatenate((np.array([self._data['x']]).T, np.array([self._data['y']]).T), axis=1)
    centers = np.concatenate((np.array([self._data['cx']]).T, np.array([self._data['cy']]).T), axis=1)
    for label in range(num_unique_labels):
        p = [p for i, p in enumerate(points) if self._data['labels'][i] == label]
        imgs = [img for i, img in enumerate(self._data['images']) if self._data['labels'][i] == label]
        self._result['clusters'].append({
            'images': imgs,
            'label': label,
            'points': p,
            'center': centers[label],
        })

Inherited members

• Module:
  • get_module_results
  • run
  • visualize