Idea for project

Game Engine Game Engine Support and Discussion
1

Hello. By combinding object recognition from Open Cv and an overlayed blueprint you could make a visable part catalog and learn about the item or fix it teach etc. This on a tablet could make anyone an service tech

2

Another possiblity would be mixing in on board diagnostics utilities so a
brocken part could glow and blink etc

3

REALITY CHECK

  1. Computer vesion and recognition is hard. I couldn’t do it. I’m not saying you can’t, but be aware that if anything, this would be near the hardest blender project ever attempted. To do this you would need to have a good knowledge of programming to complete it. (in python for the blender end, and C++ for the openCV end)
  2. It has been done. There are quite a few applications on the net already that can identify objects. Look at buildAR goblinXNA, heck, just run a search for “augmented reality”
  3. Getting information about an object. Once you have identified the object in question you then have to find out about it. So you need a large database of information. Getting this straight from the web is hard. Is the first result in google the best? Maybe you could create your own database of PDF’s, but finding all the ones would be hard.
  4. Identifying the broken bit. Um, how? Especially since:
  • Broken tech bits are broken on the inside, ie broken resistor, faulty chip. These cannot be identified by vision, only by knowing what it’s meant to do, and testing with a multi-meter or something similar.
  • Your average tablet’s camera isn’t good enough quality to follow and compare the traces on a circuit board, let alone read the writing on a chip and get it’s specs.

So let’s say that your not doing tech, only big things that the camera can identify.
Then you have to have a 3D model of the object as a whole complete bit, match up rotation and position on screen, and compare the two to see if the real one doesn’t fit. Finding this is hard as you have to find a point on the objects surfaces position in 3D space, meaning you need 2 cameras to use trigonometry to get it’s distance away from the camera. Then allow for wear and tear, and finally report if it is “functional” or not.

So yes, it could be done (possibly), it would not be easy, would have little practical use, and would probably be unreliable.
Good luck. It would be awesome to see, and I hope that someday someone will, it would make an awesome tool.
Geoff

4

I am talking about using pre existing open source code, and starting with a tiny library maybe three devices and proving the concept, and maybee getting on the summer of code to make a full version

5

Also blue tooth powered obd with documentation in the device would make it simple