Visualizations

aitlas.visualizations.classification module

Classes and methods for visualizations for classification tasks.

plot_confusion_matrix(confusion_matrix, axes, class_label, class_names, fontsize=14)

Plots a confusion matrix.

Parameters:

• confusion_matrix (array-like of shape (n_classes, n_classes)) – The confusion matrix to plot.

• axes (matplotlib.axes.Axes) – The matplotlib axes object to plot on.

• class_label (str) – The label of the class for the confusion matrix.

• class_names (list of str) – The names of the classes.

• fontsize (int, optional) – The fontsize for the plot, defaults to 14.

plot_multilabel_confusion_matrix(cm_array, labels, dataset_name, output_file)

Plots multiple confusion matrices in a .pdf format for multilabel tasks.

Parameters:

• cm_array (list of array-like of shape (n_classes, n_classes)) – The array of confusion matrices.

• labels (list of str) – The labels for the classes.

• dataset_name (str) – The name of the dataset.

• output_file (str) – The file path to save the plot.

plot_multiclass_confusion_matrix(cm_array, labels, dataset_name, output_file)

Plots multiple confusion matrices .pdf format useful for multiclass tasks.

Parameters:

• cm_array (list of array-like of shape (n_classes, n_classes)) – The array of confusion matrices.

• labels (list of str) – The labels for the classes.

• dataset_name (str) – The name of the dataset.

• output_file (str) – The file path to save the plot.

class PrecisionRecallCurve(y_true, y_pred, y_prob, labels, file, **kwargs)

Bases: BaseDetailedVisualization

plot()

Generates and plots the precision recall curve.

Returns:

matplotlib.figure.Figure object with the plot

Return type:

matplotlib.figure.Figure

class ImageLabelsVisualization(y_true, y_pred, y_prob, labels, file, **kwargs)

Bases: BaseDetailedVisualization

Class for visualising predictions for an image.

Initialize the ImageLabelsVisualization class.

Parameters:

• y_true (array-like of shape (n_samples,)) – Ground truth (correct) labels.

• y_pred (array-like of shape (n_samples,)) – Predicted labels, as returned by a classifier.

• y_prob (list of float) – The predicted probabilities.

• labels (list of str) – The labels for the classes.

• file (str) – The file path to save the plot.

• kwargs – Additional keyword arguments.

plot()

Plots the image with the predictions.

Returns:

matplotlib.figure.Figure object with the plot

Return type:

matplotlib.figure.Figure

plot_prediction(img, probs, classes)

Display image and predictions from model

Parameters:

• img (array-like or PIL image) – Image to plot.

• prob (list of float) – The predicted probabilities.

• classes (list of str) – The labels for the classes.

Returns:

matplotlib.figure.Figure object with the plot

Return type:

matplotlib.figure.Figure

display_image_labels(image, y_true, y_pred, y_prob, labels, output_file)

precision_recall_curve(y_true, y_pred, y_prob, labels, output_file)

aitlas.visualizations.eopatch module

Method for visualising predictions in EOpatch format for multi-temporal data. Useful for croptype classification tasks.

display_eopatch_predictions(eopatches_path, patch, y_pred, test_index, y_true, classmapping)

Displays the predictions of an EOPatch.

Parameters:

• eopatches_path (str) – The path to the directory containing EOPatches.

• patch (str) – The specific patch to be displayed.

• y_pred (array-like of shape (n_samples,)) – The predicted labels, as returned by a classifier.

• test_index (pandas.DataFrame) – The indices of the test set.

• y_true (array-like of shape (n_samples,)) – Ground truth (correct) labels.

• classmapping (pandas.DataFrame) – A mapping from class labels to class names.

Returns:

matplotlib.figure.Figure object with the plot

Return type:

matplotlib.figure.Figure

aitlas.visualizations.grad_cam module

Classes and methods for GRAD-CAM visualizations, used for classification tasks.

Note

Based on the implementation at: jacobgil/pytorch-grad-cam

class ActivationsAndGradients(model, target_layers, reshape_transform)

Bases: object

Class for extracting activations and registering gradients from targetted intermediate layers.

Parameters:

• model – The model to be evaluated

• target_layers – The target layers from which to extract activations and gradients

• reshape_transform – A function to reshape the activation and gradient data

save_activation(module, input, output)

Saves an activation.

Parameters:

• module – The module from which to save the activation

• input – The input data to the module

• output – The output data from the module

save_gradient(module, input, output)

Saves the gradient.

Parameters:

• module – The module from which to save the gradient

• input – The input data to the module

• output – The output data from the module

release()

get_2d_projection(activation_batch)

scale_cam_image(cam, target_size=None)

show_cam_on_image(img, mask, use_rgb=False, colormap=2, image_weight=0.5)

This function overlays the cam mask on the image as an heatmap.

By default the heatmap is in BGR format. :param img: The base image in RGB or BGR format. :param mask: The cam mask. :param use_rgb: Whether to use an RGB or BGR heatmap, this should be set to True if ‘img’ is in RGB format. :param colormap: The OpenCV colormap to be used. :param image_weight: The final result is image_weight * img + (1-image_weight) * mask. :returns: The default image with the cam overlay.

Parameters:

• img (ndarray) –

• mask (ndarray) –

• use_rgb (bool) –

• colormap (int) –

• image_weight (float) –

Return type:

ndarray

class ClassifierOutputTarget(category)

Bases: object

reshape_transform(tensor, height=14, width=14)

class BaseCAM(model, target_layers, use_cuda=False, reshape_transform=None, compute_input_gradient=False, uses_gradients=True)

Bases: object

Parameters:

• model (Module) –

• target_layers (List[Module]) –

• use_cuda (bool) –

• reshape_transform (Callable) –

• compute_input_gradient (bool) –

• uses_gradients (bool) –

get_cam_weights(input_tensor, target_layers, targets, activations, grads)

Parameters:

• input_tensor (Tensor) –

• target_layers (List[Module]) –

• targets (List[Module]) –

• activations (Tensor) –

• grads (Tensor) –

Return type:

ndarray

get_cam_image(input_tensor, target_layer, targets, activations, grads, eigen_smooth=False)

Parameters:

• input_tensor (Tensor) –

• target_layer (Module) –

• targets (List[Module]) –

• activations (Tensor) –

• grads (Tensor) –

• eigen_smooth (bool) –

Return type:

ndarray

forward(input_tensor, targets, eigen_smooth=False)

Parameters:

• input_tensor (Tensor) –

• targets (List[Module]) –

• eigen_smooth (bool) –

Return type:

ndarray

get_target_width_height(input_tensor)

Parameters:

input_tensor (Tensor) –

Return type:

Tuple[int, int]

compute_cam_per_layer(input_tensor, targets, eigen_smooth)

Parameters:

• input_tensor (Tensor) –

• targets (List[Module]) –

• eigen_smooth (bool) –

Return type:

ndarray

aggregate_multi_layers(cam_per_target_layer)

Parameters:

cam_per_target_layer (ndarray) –

Return type:

ndarray

forward_augmentation_smoothing(input_tensor, targets, eigen_smooth=False)

Parameters:

• input_tensor (Tensor) –

• targets (List[Module]) –

• eigen_smooth (bool) –

Return type:

ndarray

class GradCAM(model, target_layers, use_cuda=False, reshape_transform=None)

Bases: BaseCAM

Gradient-weighted Class Activation Mapping (Grad-CAM) class.

Parameters:

• model – The model to be evaluated

• target_layers – The target layers from which to extract activations and gradients

• use_cuda – Whether to use CUDA for computation (default: False)

• reshape_transform – A function to reshape the activation and gradient data (default: None)

get_cam_weights(input_tensor, target_layer, target_category, activations, grads)

aitlas.visualizations.segmentation module

Classes and methods for visualizations for segmentation tasks.

class ImageMaskPredictionVisualization(y_true, y_pred, y_prob, labels, file, **kwargs)

Bases: BaseDetailedVisualization

Class for visualizing the image mask predictions.

Initialisation

Parameters:

• y_true (array-like of shape (n_samples,)) – The ground truth labels

• y_pred (array-like of shape (n_samples,)) – The predicted labels

• y_prob (list of float) – The predicted probabilities

• labels (list of str) – The class labels

• file (str) – The output file path

plot()

Plots the image mask predictions and saves the plot to the output file.

plot_segmenation(img, probs, labels)

Displays the image and the predicted segmentation masks for each label.

Parameters:

• img (array-like or PIL image) – The input image

• probs (list of float) – The predicted probabilities

• labels (list of str) – The class labels

Returns:

The figure containing the plots

Return type:

matplotlib.figure.Figure

display_image_segmentation(image, y_true, y_pred, y_prob, labels, file)

Displays the predicted segmentation masks for each label.

Parameters:

• image (array-like or PIL image) – The input image

• y_true (array-like of shape (n_samples,)) – The ground truth labels

• y_pred (array-like of shape (n_samples,)) – The predicted labels

• y_prob (list of float) – The predicted probabilities

• labels (list of str) – The class labels

• file (str) – The output file path

save_predicted_masks(y_pred, labels, base_filepath_name)

Saves the predicted masks to the specified file path.

Parameters:

• y_pred (array-like of shape (n_samples,)) – The predicted labels

• labels (list of str) – The class labels

• base_filepath_name (str) – The base file path name