4 ms·

Something human can't see is uninteresting. Fractional values of pixel color is one thing. Our human eyes have a lot of filters (hardware and software-based) b

• sp332 11y ago
  The neural network should capture all the behavior of the relevant filters as well.
  • guard-of-terra 11y ago
    I'm not sure. Human brain uses dedicated facilities for this. You won't win much by making every neural network learn stuff from scratch that can be done once, good.
• orting 11y ago
  There is a lot of stuff the human eye cant see that is very interesting. One of the challenges in medical imaging is getting accurate labelling of images. For many labellings we see large inter- and intra-observer variability. We have both the problem that humans see something that is not interesting and miss something that is interesting. I currently work on estimating emphysema extent in CT lung scans. Emphysema can be very diffuse and it is not possible to label individual pixels, so instead we try to learn the local emphysema pattern from a global label. Neural networks are interesting for this problem because the learn the features, but it is also a "problem" because the features might not make physically sense, which could make it hard to transfer the model and convince clinicians that they should use it.
  • guard-of-terra 11y ago
    For that kind of task, you might want to filter out other things. We should just be realistic. We want to take real image, except it might be tinkered with, and make neural net tell us what we see on it, except we also want it to see what we can't see, and we want it to answer as accurate as possible, except we also want short and definitive answer. We also kind of want it to admit that image always contains more than one thing, but kind of don't.