Neural network based 3D indoor location tracking on a moving ship in the middle of the ocean with ar visualization for the crew to find guests.

You are on a cruise, you have an app on your phone to order stuff (drinks, meals etc)
Once you order an indoor location system calculates your position (based signal strength on training) on the ship(x, y, z)(deck, area etc) and sends it to the crew.

The crew wears an ar glass and once your order is ready they get a realtime ar navigation to you.

It was seriously over-engineered 😀
We used the phone’s bluetooth and beacons on the ship to calculate the position based on signal strength.

  • 11
    That seriously sounds amazing
  • 1
    Holy shit
  • 9
    Were there not places on the ship where the signal strength was identical? How did crew members identify people in a crowd?
  • 2
    @AlgoRythm Triangulation for point one, i guess, no clue about two tho.
  • 7
    Mesh networks, rfid or otherwise make this problem a lot easier to solve. With three points you can get within a few inches of the location. We did something similar at JC Penney. If you're holding a product there, the cameras are tracking you.
  • 5
    If you'd seen the security tapes I've seen, you'd never wear clothes again. 😝
  • 1
    @SortOfTested Wait why? Not that i like clothing either way, but what?
  • 2
    @AlgoRythm Yepp, triangulation.
    As for finding people in a crowd we used facial recignition.
    We implemented a paper that allowed us to do good enough facial recognition from just one pictue. The guests had a profile pic associated with their app/profile so we just needed to run the algo over a db.
    The ar glasses feeded the video into a server and the server sent back the bounding box of the guest’s face if found.
    Thus you can project a circle in AR around the person’s face you looking for.
    Also the crew got the guest’s profile pic as a fallback.
  • 3
    Ps if i remember correctly we used something based on dlib for the face recognition. Not sure tho, my teammate worked on that part. My part was mainly visualization, AR localisation, testing/validation (coming up for challanges for the system and how to solve them) and i helped in the beacon based tracking’s code a bit.
    Also i didn’t expected this many people to be intrested in this :D i might be able to get a video up to youtube if you promise not to share it too much 😀
  • 0
    Sounds pretty futuristic OwO
  • 0
    Proving Angles Congruent Two angles are congruent if they have the same measure. You already know that when two lines intersect the vertical angles formed are congruent.
  • -1
    I do not think so
Add Comment