One Wave to Explain Them All: A Unifying Perspective on Feature Attribution

Supplementary data for the paper "One Wave to Explain Them All: A Unifying Perspective on Feature Attribution" (Anonymous sumission) This folder contains the data and material to replicate our results. ** `datasets` contain the data for sounds (ESC50) and 3D shapes (3DMNIST) ** `models` contain the weights of the following pretrained models: ** `VoxelModel`: the model classifying the 3D shapes of MNIST, ** `best_full_model_Fold1_K100.pth`: the audio classifier, ** `adv`, `adv_free` and `fast_adv`: the robust ResNet50 for the robustness experiment ** `benchmark`: contains 1000 imagenet images to replicate the evaluation results for the 2D WAM.