JML_XAI_Project

About JML_XAI_Project

The JML_XAI_Project package implements the explainable AI methods SHAP and LIME for image inputs. The project was developed as part of the "Julia for Machine Learning" course at TU Berlin.

Here you can find the documentation.

Installation

You can install LIME by adding it directly from our GitHub repository. Here are the steps:

1. Open Julia's REPL (the Julia command-line interface).

2. Press ] to enter Pkg mode.

3. Run the following command to add the necessary LARS algorithm dependency:

pkg> add https://github.com/e-strauss/LARS.jl

4. Run the following command to add Lime:

pkg> add https://github.com/e-strauss/JML_XAI_Project

Usage

using ExplainableAI
using Flux
using Metalhead: ResNet
using JML_XAI_Project
using Images
using VisionHeatmaps

#usage: include("src/xai-getting-started.jl")

#Plots heatmap
    #If overlay => heatmap on image (image in black and white)
    #If blurring => heatmap blurred
    #gaussSTD = standard deviation of gauss kernel, gaussSTD higher => more blurring
function generate_heatmap(map; img=nothing, overlay=false, blurring=false, gaussSTD=2)
    map = heatmap(map.val)

    if blurring == true
        gaussKern2 = ImageFiltering.KernelFactors.gaussian((gaussSTD,gaussSTD))
        map = ImageFiltering.imfilter(map, gaussKern2)
    end

    if overlay == true
        map = (0.5.*Gray.(img) + 0.5.*map)
    end

    return map
end

img = load("data/n01742172_boa_constrictor.JPEG")
display(img)
imgVec = permutedims(channelview(img),(3,2,1))
imgVec = reshape(imgVec, size(imgVec)..., 1)
input = Float32.(imgVec)
@info size(input)

model = ResNet(18; pretrain = true);

#Explanation Using LIME
model = model.layers;analyzer = LIME(model)

#Explanation Using SHAP with Model-Agnostic Kernel (lasso has to be set to false)
#model = model.layers;analyzer = LIME(model, agnostic_kernel, false)

expl = analyze(input, analyzer);
print("Label: ", argmax(expl.output[:,1]) - 1)
heat = generate_heatmap(expl, img=img, overlay=true, blurring=true)
display(heat)

#Save Heatmap to File
#save("C:/Users/USERNAME/Desktop/heatmap boa.png", heat)