.. note::
    :class: sphx-glr-download-link-note

    Click :ref:`here <sphx_glr_download_sphinx_gallery_auto_transparency_xmpl_transparency_lime_image.py>` to download the full example code or run this example in your browser via Binder
.. rst-class:: sphx-glr-example-title

.. _sphx_glr_sphinx_gallery_auto_transparency_xmpl_transparency_lime_image.py:


==========================
Using LIME Image Explainer
==========================

This example illustrates how to use the LIME image explainer to explain a
prediction.

This example shows how to use the image LIME implementation --
:class:`fatf.transparency.predictions.surrogate_image_explainers.ImageBlimeyLime`
-- to explain a prediction of a black-box probabilistic model.


.. rst-class:: sphx-glr-horizontal


    *

      .. image:: /sphinx_gallery_auto/transparency/images/sphx_glr_xmpl_transparency_lime_image_001.png
            :class: sphx-glr-multi-img

    *

      .. image:: /sphinx_gallery_auto/transparency/images/sphx_glr_xmpl_transparency_lime_image_002.png
            :class: sphx-glr-multi-img


.. rst-class:: sphx-glr-script-out

 Out:

 .. code-block:: none

    {'bottom-left': 0.059825904043669643,
     'bottom-right': -0.27639956335921306,
     'top-left': 0.4983489307756843,
     'top-right': 0.04997774025460273}




|


.. code-block:: default

    # Author: Kacper Sokol <k.sokol@bristol.ac.uk>
    # License: new BSD

    from pprint import pprint

    import matplotlib.pyplot as plt
    import numpy as np

    import fatf

    import fatf.utils.models as fatf_models

    import fatf.transparency.predictions.surrogate_image_explainers as fatf_exp

    import fatf.vis.lime as fatf_vis_lime

    print(__doc__)

    # Fix random seed
    fatf.setup_random_seed(42)

    # Create a simple data set
    r, g, b, k = [255, 0, 0], [0, 255, 0], [0, 0, 255], [0, 0, 0]
    X = np.array(
        [
            [[r, g], [b, k]],
            [[r, b], [g, k]],
            [[r, k], [b, g]],
            [[k, g], [b, r]],
            [[k, b], [g, r]],
            [[g, k], [b, r]]
        ],
        dtype=np.uint8
    )
    y = np.array([0, 0, 0, 1, 1, 1])

    feature_names = {
      'Segment #1': 'top-left',
      'Segment #2': 'top-right',
      'Segment #3': 'bottom-left',
      'Segment #4': 'bottom-right'
    }
    class_names = {0: 'top-left-red', 1: 'bottom-right-red'}

    # Create a model that operates of 3-D images
    class KNNimg(fatf_models.KNN):
        def fit(self, X, y):
            X_r_flat = np.array([i[:, :, 0].flatten() for i in X], dtype=np.uint8)
            super().fit(X_r_flat, y)

        def predict(self, X):
            X_r_flat = np.array([i[:, :, 0].flatten() for i in X], dtype=np.uint8)
            return super().predict(X_r_flat)

        def predict_proba(self, X):
            X_r_flat = np.array([i[:, :, 0].flatten() for i in X], dtype=np.uint8)
            return super().predict_proba(X_r_flat)

    # Train a model
    clf = KNNimg(k=1)
    clf.fit(X, y)

    # Choose an instance to be explained
    image = X[0]
    image_class = 0

    # Display the instance
    plt.axis('off')
    plt.imshow(image)

    # Create a LIME image explainer
    lime = fatf_exp.ImageBlimeyLime(
        image,
        clf,
        ratio=1,
        kernel_size=1.05,
        max_dist=10,
        colour='random')

    # Explain an instance
    lime_explanation_ = lime.explain_instance(explained_class=image_class)

    # Remap the explanation
    lime_explanation = {feature_names[i]: j for i, j in lime_explanation_.items()}

    # Display the textual explanation
    pprint(lime_explanation)

    # Plot the explanation
    fatf_vis_lime.plot_lime(lime_explanation)


.. rst-class:: sphx-glr-timing

   **Total running time of the script:** ( 0 minutes  0.216 seconds)


.. _sphx_glr_download_sphinx_gallery_auto_transparency_xmpl_transparency_lime_image.py:


.. only :: html

 .. container:: sphx-glr-footer
    :class: sphx-glr-footer-example


  .. container:: binder-badge

    .. image:: https://mybinder.org/badge_logo.svg
      :target: https://mybinder.org/v2/gh/fat-forensics/fat-forensics-doc/master?filepath=notebooks/sphinx_gallery_auto/transparency/xmpl_transparency_lime_image.ipynb
      :width: 150 px


  .. container:: sphx-glr-download

     :download:`Download Python source code: xmpl_transparency_lime_image.py <xmpl_transparency_lime_image.py>`



  .. container:: sphx-glr-download

     :download:`Download Jupyter notebook: xmpl_transparency_lime_image.ipynb <xmpl_transparency_lime_image.ipynb>`


.. only:: html

 .. rst-class:: sphx-glr-signature

    `Gallery generated by Sphinx-Gallery <https://sphinx-gallery.readthedocs.io>`_