Unexpected keyword argument 'progress_bar' for lime_image.LimeImageExplainer().explain_instance()

[momchilgeorgiev]

Similar to issue #660 and #480 (specifically this part), lime_image.LimeImageExplainer().explain_instance() throws out an error when trying to hide tqdm progress bar.

This is for lime 0.2.0.1 installed via pip. I made sure to update before posting here. I am also running everything in wsl2, config: Ubuntu 22.04.3 LTS in a conda environment, because of TensorFlow requirements on windows systems.

I wanted to share since sometimes it's pretty unusable when working in a notebook. Am I doing something wrong?

I have also checked the source code:

Code:

def plot_lime(img_path, model):
    # img_path - str
    # model - tf model instance
    # Get an image path and a model and plot 

    print(f'Processing {img_path}')
    # Explain a prediction
    explainer = lime_image.LimeImageExplainer()
    segmenter = SegmentationAlgorithm('slic', n_segments=100, compactness=1, sigma=1) 

    img = preprocess_image(img_path)[0]  # Preprocess the image

    # Make predictions
    preds = model.predict(img[np.newaxis, ...])  # Add batch dimension
    top_pred_index = np.argmax(preds[0])  # Index of the top prediction
    top_pred_label = ['NORMAL', 'PNEUMONIA'][top_pred_index]  
    top_pred_prob = preds[0][top_pred_index]  # Probability of top prediction

    # Get the explanation
    explanation = explainer.explain_instance(img.astype('double'), 
                                            classifier_fn=model.predict, 
                                            top_labels=1, 
                                            hide_color=0, 
                                            num_samples=1000, 
                                            segmentation_fn=segmenter,
                                            progress_bar=False
                                            )

    # Display the top label's explanation
    temp, mask = explanation.get_image_and_mask(explanation.top_labels[0], positive_only=True, num_features=50)
    plt.imshow(mark_boundaries(temp / 2 + 0.5, mask))
    plt.show()

    print(f"Model's predicted class: {top_pred_label} with probability {top_pred_prob}")

Call function:

plot_lime(image_path, resnet_balanced)

Error:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[29], [line 1](vscode-notebook-cell:?execution_count=29&line=1)
----> [1](vscode-notebook-cell:?execution_count=29&line=1) plot_lime(image_path, resnet_balanced)

Cell In[28], [line 27](vscode-notebook-cell:?execution_count=28&line=27)
     [24](vscode-notebook-cell:?execution_count=28&line=24) top_pred_prob = preds[0][top_pred_index]  # Probability of top prediction
     [26](vscode-notebook-cell:?execution_count=28&line=26) # Get the explanation
---> [27](vscode-notebook-cell:?execution_count=28&line=27) explanation = explainer.explain_instance(img.astype('double'), 
     [28](vscode-notebook-cell:?execution_count=28&line=28)                                         classifier_fn=model.predict, 
     [29](vscode-notebook-cell:?execution_count=28&line=29)                                         top_labels=1, 
     [30](vscode-notebook-cell:?execution_count=28&line=30)                                         hide_color=0, 
     [31](vscode-notebook-cell:?execution_count=28&line=31)                                         num_samples=1000, 
     [32](vscode-notebook-cell:?execution_count=28&line=32)                                         segmentation_fn=segmenter,
     [33](vscode-notebook-cell:?execution_count=28&line=33)                                         progress_bar=False
     [34](vscode-notebook-cell:?execution_count=28&line=34)                                         )
     [36](vscode-notebook-cell:?execution_count=28&line=36) # Display the top label's explanation
     [38](vscode-notebook-cell:?execution_count=28&line=38) temp, mask = explanation.get_image_and_mask(explanation.top_labels[0], positive_only=True, num_features=50)

TypeError: explain_instance() got an unexpected keyword argument 'progress_bar'