SHAX-PC : Adapting XAI methodology SHAP for point cloud data

This research created a custom SHAP framework designed to explain a PointNet model’s prediction. By calculating the marginal contribution of an individual point with respect to a coalition towards the model’s prediction. Individual points’ SHAP value shows the level of contribution that point has towards the classification of a PointNet Model in the Model Net 10 dataset. This research successfully implements the custom SHAP framework and showcases which points contribute to the classification of a chair.