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.
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
