Surrogate Model-Based Explainability Methods for Point Cloud NNs

This work is based on point cloud networks (this repo) and local surrogate model-based explainability method LIME (this repo). Please follow their instructions to build up the environments.

Usage

After installing LIME, copy the scripts in lime_script/ into lime/ (lime_base will be replaced)

Move modelnet40_test_lime.txt to data/modelnet40_normal_resampled/ or run sample_test_data.py to sample user-defined test set (this py file should also be moved into data/modelnet40_normal_resampled/).

Visualize the explanation for one instance:

python LIME_single.py

Evaluate a batch of explanations:

python test_batch_LIME.py