PartitionShap demo: channel-wise explanations of raster-based machine learning models

Exploring PartitionShap on raster datasets of an arbitrary number of bands/channels

These notebooks use our SHAP fork with modifications to make it easier to use and visualize multi-channel explanations. The EuroSAT notebooks use TorchSat, a library for working with PyTorch models whose inputs are rasters with arbitrary number of channels.

Citation

@inproceedings{krell2022explaining,
  title={Explaining Complex 3D Atmospheric CNNs Using SHAP-Based Channel-wise XAI Techniques with Interactive 3D Visualization},
  author={Krell, Evan and Kamangir, Hamid and Friesen, Josh and Judge, Julianna and Collins, Waylon G and King, Scott Alan and Tissot, Philippe E},
  booktitle={102nd American Meteorological Society Annual Meeting},
  year={2022},
  organization={AMS}
}

Notebooks

1. PartitionSHAP: ImageNet (RGB) demo
2. PartitionSHAP: EuroSAT (RGB) demo
   • pretrained weights: https://drive.google.com/file/d/14BYvrjem4dbmkmibmmSD2cDq2ZvStfVG/view?usp=sharing
3. PartitionSHAP: EuroSAT (13-band) demo
   • pretrained weights: https://drive.google.com/file/d/1gYtOpYdMCCMxCkEhRzK1dDyE6tOMIOKa/view?usp=sharing
4. PartitionSHAP: Tornado (4-band) demo

3D SHAP viewer tool (prototype)

A simple prototype of the viewer is included for completeness. Further development of this tool occurs in the xai-raster-vis-tools repo.

python SHAP_3D_viewer.py --help
Usage: SHAP_3D_viewer.py [options]

Options:
  -h, --help            show this help message and exit
  -f FILE, --file=FILE  Path to 3D SHAP values (.npz)
  -d DATA_NAME, --data_name=DATA_NAME
                        Name of SHAP values in the input SHAP values (.npz)
                        file.
  -e, --show_edges      Show edges of grid elements

Todo

• Evan: Use test instead of validation data for XAI runs in notebooks -> not currently repoducable since will choose a random validation case
• Josh: Add colab notebook with RGB & RGB+NIR model train, test, PartitionShap results (other XAI not is not for this repo!)