Insightful Saliency Maps
Make your own stylized image. See what the AI predicts. Is it biased towards texture or shape?
Explore the saliency maps of stylized images. See how they compare to the saliency map of the original, un-stylized image.
A saliency map identifies what regions in an image are the most important to a neural network's prediciton. XRAI is a saliency method that uses regions and colors to interpret how a computer understands an image. Typically, the image is split into a gradient of colors that represent what features in the image are most important to identifying the class of the image all the way to what features are least important to identifying the class of the image.
1 Choose a stylized image:
Prediction: ; % confidence
Original image:
Prediction: ; % confidence
Compute the average saliency map
1 Choose an Image Class:
2 Select the images below to include in the average:
Does your perception align with AI?
Color in the most to least important regions of the dog below to see how you compare to Explainable AI.
The AI makes predictions based on regions it feels are important to identifying the image. What regions of the image do you think are most important to identifying this as a Doberman Pinscher? What regions are important and somewhat important? Finally, what regions are least important?
Select one of the color buttons below to start drawing on the image which regions are most and least important.
Select a color
Percentage of pixels aligned with the AI so far
= This is how the AI sees
Here's what the AI sees
Compare accuracy of different models through saliency maps. Is there one that consistently performs better than others?
1Select models to compare
Comparison method: Intersect XRAI
2Select stylized image
Model Comparison:
