• ayyy@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    9
    ·
    edit-2
    19 hours ago

    Please do not consume plants based on a visual ID by a convolutional neural network. It’s a dangerous and fundamentally flawed approach. Many plants can only be properly identified by observing specific parts that don’t turn out from a single angle photo, such as formations on the underside of leaves, or a specific curl direction of a flower stamen etc. etc.

    • Danquebec@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      1
      ·
      10 hours ago

      The app allows you to take several photos, label those parts (“leaves”, “flower”, etc.), and then you get several results, sorted by correspondence according to the neural network. When you click on a result, you get photos of different parts of the plant, allowing you to compare to your plant and judge if they’re the same.

      • pseudo@jlai.lu
        link
        fedilink
        English
        arrow-up
        2
        ·
        7 hours ago

        In now this app and I use it but I agree with the previous comment. PlantNet should be an aid. One among many clue helping you getting the ability to recognis on your own which plant it is. Then we you don’t need PlantNet to recognise the plant 200 % and to say “It can only be this plant because of Z, of Y and also of X”, only then you should eat it.

        Otherwise, your playing with fire. That is your health and potentially your life.

    • Screamium@lemmy.world
      link
      fedilink
      English
      arrow-up
      3
      ·
      13 hours ago

      Well, yeah if you want to eat something you want to be sure. Even if you only take one photo, you can look at pictures of the leaves, flowers, and fruit. There’s a link to other sources, like Wikipedia, where you can get more info.