Tuesday, March 17, 2015

Here's a bajillion dollars, if you can spend a few million on software development.

Here's a bajillion dollar idea for someone who can make it happen.  Combine image recognition software + 3d modelling software + time & location tagged images (and videos) to build a 3d virtual world, instanced by time.  Conceptually, with 2 or more photos of the same building from different angles, one could start to build a 3D model of that building.  The more photos you use, the better the model could be. Those 2 or more photos could even be separate frames from a video or movie.  So if you knew where a video was shot and when, you might be able to build a 3D model of the set of the video.  Then, you could use subsequent photos or video to refine the model, or build a separate model of how that place appeared at a different time.  To start with, you'd have to go with popular places that are frequently photographed, like Times Square, and use all of the photos you can find to build a model of Times Square, and instance it for different times.  Eventually, you might be able to take a virtual tour of Times Square in 1950, or 1970 or any date you please.

As more and more data is added to a given model, one might even be able to work backwards and identify the time and location of a random photo, if it includes a popular landmark.

And when Google decides to take all of their stored photos and videos, and run it through this system, you might be able to take a virtual reality stroll through nearly any city for the last 40-50 years, or longer ago for really well photographed places.

And once they run facial recognition software on their archive, and scan it for license plates, they might be able to track any heavily photographed event (or person) from years ago.  For instance they could combine photos and in game footage of a Yankees game, and be able to virtually put you in the stands for any game of the last 20-30 years, and watch it play out from that viewpoint, and maybe even be able to accurately model and identify the people sitting next to you.  Of course, it wouldn't just be sporting events that could be recreated, using this technique, one would be able to tour the World Trade Center for instance or walk the decks of the Titanic.

And that was my daydream on my drive in to work this morning.


  1. Fucking 'ell.

    Second time is the charm.

    Fascinating, yet terrifying at the same time, yet very much a plausibility. In fact I am almost certain that we are in the beta stage of development.

    Values for "we" equalling nefarious government or corporate bastardy.

    Nice one, happy pattys and all that, email is coming your way.

  2. I guess it's an idea... But I _already_ strolled through a shitload of cities over the last forty or fifty years. Most of them I have no desire to ever see again. Hey I can't help it - woke up with my curmudgeon hat on this morning. I'll just note that the sat-nav in my new car has 3D buildings. Pretty cool and actually helpful in downtown D.

    OT, I got an itch for a new bike. That GSXR front end on my trusty SV1000 is fucking killing me. Quick lesson in how to make a bike unrideable for an old fart: start with a bike with notoriously bad ergos - an SV1000S, say - then put a goddamn GSXR front end on to make the WORST. ERGOS. EVER. Soon as my ribs have healed sufficiently Imma check out a KTM 990 SM and maybe a Ducati Hyperstrada.