Neural nets based decision trees are so 90s.
And your tessellation is pretty slow and pretty low res. Have you heard of multires? Also if you used wavelets you probably could get a higher framerate with more resolution and less processing and better registration.
For example and this is not advanced by any means, this is a 3d reconstruction based only on remote GIS sensing that I developed for a company a while back.
It's all derived from different data sources using a lot of parallel processing using GPU cores, making sure the data sets are fully registered.
Not real time but it's doing a lot and it can find trees, houses, concrete, water and map it on DEM data. It can even tell you what kind of tree it is or what type of material a road is.
And I'm not that great either, others have done far better than me. As for neural nets I actually have an SPIE paper published in 1992 and I was a voting member in the formation of comp.ai.neural-nets.natural-lang usenet channel way back in the early 90s. I probably forgot more things that you know.
Your problem is that you are self taught, because that is pretty obvious, so you don't have a common language with the field, so you'll get a lot of blank stares because you are not equipped to describe what you are doing. Not because what you are doing is complicated, just your way to communicate is what is complicated.
And I wouldn't pipe in at all and let you on your ego narcissistic crusade but you just happened to insult
https://en.wikipedia.org/wiki/Grace_HopperAs for the Kinect, meh, what's so difficult in projecting something once you have the z depth already and using someone else's software, because I'm pretty sure you are not computing the time of flight yourself. Now that would be impressive if you could actually get the timings right, but for that you might need to know the length of a nanosecond and a bunch of optics and physics.
Look at Microsoft's HoloLens, now that is impressive. Although I do like CastAR but that requires the retro reflective surface. ZCam's tech (in the kinect because MS bought them) is cool and all but it doesn't work well in the sun.
Maybe come up with something better like some spread spectrum based time of flight so you can use it outdoors regardless of other IR sources.
And yeah, making an undocumented video signal format work is more challenging that what you are doing, because time has to be just right and it's not like you can just slap a bunch of random code together.
Weeding out nonsense is not holding progress back, because doing the weeding out prevents others from wasting time and resources so we can have real progress.
It's a perspective thing. And my perspective on your shirtless videos smoking and trying to be cool, actually corroborates what I mentioned on my previous post. Don't get me wrong I used to do the same on my 20s (other than posting on youtube) and weed made everything make sense and had a lot of stupid philosophical and pseudo science conversations with my peers, but then I grew up
I still remember one about trying to convince others that at a quantum level a pencil could have consciousness.
Edit: and by the way I did convince them, but I guess they just got tired to listen to my rambles instead.
Home
Help
Search
Login
Register
