Plots

Functions used to visualize the patch/embeddings/results.

plot_embeddings(input_dictionary, embedding_key, annotation_keys, sup_title=None, n_col=3, cmap='inferno')[source]

Takes a dictionary with embeddings and multiple annotations and make a multi-panel figure with each panel showing one annotation.

Parameters

  • input_dictionary (dict) – dictionary with input data

  • embedding_key (str) – str corresponding to the embeddings in input_dictionary. Embedding have shape (n_sample, latent_dim). Only the first two latent dimensions will be used for plotting.

  • annotation_keys (List[str]) – List[str] corresponding to annotations in input_dictionary.

  • sup_title (Optional[str]) – the title (if any) for the figure

  • n_col (int) – how many columns to have in the multi-panel figure

  • cmap (Colormap) – the color map to use for the continuous variable. The categorical variable will have a different cmap.

Returns

fig – multi-panel figure with the embeddings

Return type

Figure

pad_and_crop_and_stack(x, pad_value=0.0)[source]

Takes a list of tensor and returns a single batched_tensor. It is useful for visualization.

Parameters

  • x (List[Tensor]) – a list of tensor with the same channel dimension but possibly different width and heigth

  • pad_value (float) – float, the value used in padding the images. Defaults to padding with black colors

Returns

tensor – A single batch tensor of shape \((B, c, \text{width}_\text{max}, \text{height}_\text{max})\)

Return type

Tensor

show_batch(tensor, cmap=None, n_col=4, n_padding=10, title=None, pad_value=1, normalize=True, normalize_range=None, figsize=None)[source]

Visualize a torch tensor of shape: \((*, \text{ch}, \text{width}, \text{height})\). It works for any number of leading dimensions.

Parameters

  • tensor (Tensor) – the torch.Tensor to plot

  • cmap (Optional[str]) – the color map to use. If None, it defaults to ‘gray’ for 1 channel images, RGB for 3 channels images and ‘tab20’ for images with more that 3 channels.

  • n_col (int) – int, number of columns in the image grid

  • n_padding (int) – int, padding between images in the grid

  • title (Optional[str]) – str, the tile on the image

  • pad_value (int) – float, pad_value

  • normalize (bool) – bool, if tru normalize the tensor in normalize_range

  • normalize_range (Optional[Tuple[float, float]]) – tuple, if not specified it is set to (min_image, max_image)

  • figsize (Optional[Tuple[float, float]]) – size of the figure

Return type

Figure

show_raw_all_channels(data, n_col=4, cmap=None, figsize=None, titles=None, sup_title=None, show_colorbar=None, legend_colorbar=None, show_axis=True)[source]

Visualize a torch tensor of shape: \((*, \text{ch}, \text{width}, \text{height})\) or a list of tensors of shape \((\text{ch}, \text{width}, \text{height})\).

Parameters

  • data (Union[Tensor, List[Tensor]]) – A torch.tensor of shape \((*, \text{ch}, \text{width}, \text{height})\) or a list of tensor of shape \((\text{ch}, \text{width}, \text{height})\)

  • n_col (int) – number of columns to plot the data

  • cmap (Optional[str]) – the matplotlib color map to use. Defaults to RBG (if data has 3 channels) or ‘tab20’ otherwise.

  • figsize (Optional[Tuple[float, float]]) – Optional, the tuple with the width and height of the rendered figure

  • titles (Optional[List[str]]) – list with the titles for each panel

  • sup_title (Optional[str]) – the title for the entire image

  • show_colorbar (Optional[bool]) – bool, if yes show the color bar

  • legend_colorbar (Optional[List[str]]) – legend for the colorbar

  • show_axis (bool) – If True (default) show the axis

Returns

fig – A figure with * panels each panel is a rendering of a tensors of shape \((\text{ch}, \text{width}, \text{height})\)

Return type

Figure

show_raw_one_channel(data, n_col=4, cmap=None, in_range='image', scale_each=True, figsize=None, titles=None, sup_title=None, show_axis=True)[source]

Visualize a torch tensor of shape \((*, \text{width}, \text{height})\). or a list of tensor of shape \((\text{width}, \text{height})\). Each leading dimension is shown separately.

Parameters

  • data (Union[Tensor, List[Tensor]]) – A torch.tensor of shape \((*, \text{width}, \text{height})\) or list of tensor of shape \((\text{width}, \text{height})\).

  • n_col (int) – number of columns to plot the data

  • cmap (Optional[str]) – the matplotlib color map to use. If None use ‘gray’ colormap.

  • in_range (Union[str, Tuple[float, float]]) – Either a tuple specifying a a min and max value (for clamping) or a string ‘image’. If ‘image’ the min and max value are computed form the image itself. Value are clamped in_range and then transformed to range (0.0, 1.0) for visualization.

  • scale_each (bool) – bool if true each leading dimension is scaled by itself. It has effect only if in_range = ‘image’

  • figsize (Optional[Tuple[float, float]]) – Optional, the tuple with the width and height of the rendered figure

  • titles (Optional[List[str]]) – list with the titles for each small image

  • sup_title (Optional[str]) – str, the title for the entire image

  • show_axis (bool) – If True (defaults) show the axis.

Returns

fig – A figure with (*) panels. Each panel is a rendering of a tensor of shape \((\text{width}, \text{height})\)

Return type

Figure

plot_cdf_pdf(pdf_y=None, cdf_y=None, x_label=None, sup_title=None)[source]

Plot the Probability Density Function (pdf) and Cumulative Density Function (cdf).

Parameters

  • pdf_y (Union[ndarray, Tensor, None]) – array like data

  • cdf_y (Union[ndarray, Tensor, None]) – Optional. The cdf to plot. If not present it can be computed internally from pdf_y

  • x_label (Optional[str]) – the label of the x-axis

  • sup_title (Optional[str]) – the title for both panels

Returns

fig – A two panel figure with the pdf and cdf.

Return type

Figure

plot_clusters_annotations(input_dictionary, cluster_key, annotation_keys, titles=None, sup_title=None, n_col=3, figsize=None)[source]

ADD DOC STRING

Return type

Figure

plot_multiple_barplots(data, x, ys, n_col=4, figsize=None, y_labels=None, x_labels_rotation=90, x_labels=None, titles=None, y_lims=None, **kargs)[source]

IMPROVE DOCSTRING

Takes a dataframe and make multiple bar plots leveraging seaborn.barplot(y=y, x=x, data=data)

Parameters

  • data (DataFrame) – a dataframe with the data to plot

  • x (str) – names of (independent) variables in data

  • ys (List[str]) – names of (dependent) variables in data

  • n_col (int) – number of columns panels in the figure

  • figsize (Optional[Tuple[float, float]]) – size of the output figure

  • x_labels (Optional[List[str]]) – label for the x-axis

  • y_labels (Optional[List[str]]) – labels for the y-axis

  • x_labels_rotation (int) – rotation in degree of the x_labels (default 90)

  • titles (Optional[List[str]]) – titles for each panel

  • y_lims (Optional[Tuple[float, float]]) – set limits for the y coordinate for all the panels

  • kargs – any argument passed to seaborn.barplot such as hue,

Returns

fig – Figure with XXX panels

Return type

Figure

plot_knn_examples(input_dict, embedding_key, image_key, n_neighbors=3, examples=6, metric='euclidean', cmap=matplotlib.pyplot.cm.viridis, figsize=None, plot_histogram=False, max_distance=None, **kargs)[source]

NEED TO BE REARRANGED b/c INTERFACE HAS CHANGED

Parameters

  • input_dict (dict) – dictionary with the images of the crops and their embeddings

  • embedding_key (str) – str, key corresponding to the embeddings in the input_dict (or distance_matrix if :attr:’metric’ == ‘precomputed’)

  • image_key (str) – str, key corresponding to the images in the input_dict

  • n_neighbors (int) – int, how many nearest neighbour to plot

  • examples (Union[int, List[int]]) – either a list of int with the index of the example whose nearest neighbour to show or an int. If an int, the example will be selected at random.

  • metric (str) – which metric to use. Either “euclidean”, “cosine” or “contrastive”

  • cmap (ListedColormap) – the colormap to use for plotting

  • figsize (Optional[Tuple[int, int]]) – size of the image. Default to None

  • plot_histogram (bool) – if true, the histogram of distances is shown

  • max_distance (Optional[float]) – float, if given distances larger than this value will be excluded from the histogram.

  • kargs – additional parameter which will be passed to matplotlib.hist() function. For example ‘bins=50’, ‘density=True’.

Returns

a matplotlib figure with example fo knn neighbours.