Visualization: Saliency Maps :: Ori Codes

[utterances-bot]

Visualization: Saliency Maps :: Ori Codes

Ori's self-driving RC car

https://ori.codes/artificial-intelligence/visualization/saliency-maps/

[gcordova19]

Hi,

I run your code but I didn't know why you use
upscaled_activation = np.ones((53, 73))

Checking your inspiration github i realize
upscaled_activation = np.ones((3, 6))

So I have problems
averaged_activation = np.mean(activations[layer], axis=3).squeeze(axis=0) * upscaled_activation
In my case np.mean(activations[layer], axis=3).squeeze(axis=0) is (1, 18)
if I multiply by upscaled_activation in any case work by dimentions.

I would like to know why to use that dimension in upscaled_activation