pl-covidnet-grad-cam

Table of Contents

Abstract

ChRIS Plugin for Explainable AI visualization using the Grad-CAM algorithm

Synopsis

python grad_cam.py                           \
    [-h] [--help]                            \
    [--man]                                  \
    [--meta]                                 \
    [-v <level>] [--verbosity <level>]       \
    [--version]                              \
    [--modelname <modelname>]                \
    --imagefile <imagefile>                  \
    --predmatrix <predmatrix>                \
    <inputDir>                               \
    <outputDir>

Arguments

[-h] [--help]
If specified, show help message and exit.

[--man]
If specified, print (this) man page and exit.

[--meta]
If specified, print plugin meta data and exit.

[-v <level>] [--verbosity <level>]
Verbosity level for app. Not used currently.

[--version]
If specified, print version number and exit.

[--modelname]
The name of the model being used, this is optional (default is COVIDNet-CXR4-B).

[--imagefile]
The name of the input image in the input directory, this is required

[--predmatrix]
The name of the prediction matrix file in the input directory, this is required

Local Build

DOCKER_BUILDKIT=1 docker build -t local/pl-covidnet-grad-cam .

Run

docker run --rm -v $PWD/in:/incoming -v $PWD/out:/outgoing                       \
    darwinai/covidnet-grad-cam-pl covidnet-grad-cam                              \
        --imagefile ex-covid.jpg --predmatrix raw-prediction-matrix-default.json \
        /incoming /outgoing

Models

The COVIDNet-CXR4-B model is downloaded from https://drive.google.com/drive/folders/1i5XxVy6A6Dwn0IIoGqpbvQo3xgWlVgB_ For more information, visit https://github.com/lindawangg/COVID-Net/blob/master/docs/models.md

Note

Grad-CAM largely depends on the provided reference model, so make sure that the model that is used to determine the result that is used as input exactly matches the provided reference model.

Acknowledgement

Insik Kim(insikk) for initial Grad-CAM implementation for ResNet and VGG using tensorflow: https://github.com/insikk/Grad-CAM-tensorflow