3 ms·

I disagree. Information is presented to humans visually. If your models can process visual information, it is the most robust of possible solutions.