I guess blurring it is similar to reducing the resolution or to looking at the image from further away. It's interesting that the neural net figures out the ci

I guess blurring it is similar to reducing the resolution or to looking at the image from further away. It's interesting that the neural net figures out the circles, but not the words. Because the circles are also not so easily apparent from looking closely at the image. It could also be whirly lines.