Your RC car videos were better and more to my taste, this, I'm afraid is not

Fantastic effort 🎉

One of these days bro is casually going to solve one of the biggest problem of the computer vision industry.

All I am is a web designer that only knows typescript and some C# mock me relentlessly.

Very cool!

🐐

gongsterr

Good you did not take any OS course, since visionOS is, actually, not an OS.

Another Banger

HE'S BACK!

Nice work Steven! I recognized that cameo 😂

amazing hardwork as always!

Pet peeve: Lighthouse is inside-out. The sensors are on the HMD. An outside-in tracking system would have the sensors (or cameras) in the world looking at the headset, like the original Oculus Rift.

Your animation stuff has gotten insanely good

love the video!

very nice vid

really cool! very @StuffMadeHere vibe, which is great

That’s incredible and I loved the process! Excited to see the next one!

Woah. You bloody maso. That was dank!

Nice video, Why don't you use depth model estimation.

Getting this kind of result with just 2 cameras is impressive

What if you... like... were to add more cameras ? would the accuracy be greater ? Although, I can imagine it would require more processing power, since now you have to compare more frames together (camera1 + camera2, camera2+camera3, etc ....)

Amazing that you shared your learning experience with us!

It's always such a pleasure to see a software developer lose their absolute mind on something

You could sell an online course "writing SLAM from scratch"

Are you going to continue your project, to compete with ORB_SLAM, RTABmap, etc?

Steven finally dropped a new video! 🎉

Hey if you're good at slam and robotics maybe you should make a robot lawn mower that doesn't suck and isn't super proprietary like all the others.

Not to be pedantic, but it is: He lent me not: He borrowed me. that is even the case for the entirety of Florida and Cincinnati

What a madlad, trying to build a good SLAM implementation from scratch. Regardless of your results in that realm, well done! The explanations are excellent. I think you would have had an easier time getting some basic 3D LiDAR (some of the mems based ones are getting cheap) and doing something like ICP with a pre-scan and using that to localize your cameras. Future video idea I would really appreciate if you want to go down another rabbit hole: using SLAM with 2+ "headsets" and communicating between the two so that they know where they are and where each other are.

Are you going to make like a quest 3 app? I’m kinda confused what the end goal is

Absolutely amazing

我估计你是想给f1tenth做visual slam然后发现Jetson上装orb slam基本不可能 完了就想自己写一个是吧哈哈哈

Like yr maxim

awesome!

My VR headset is like 7 years old and the pixels per inch isn't high enough to read anything that's not right in front of your face, so it's not realistic for work. The thing I want to see the most about the vision pro is the high 3k ppi! wow!

5:59 isn't it just glue??

Great video. Brings back memories of my senior project

NICE

Great video!!!

Wow this was fire, the editing is beautiful

this is so cool! i've been waiting for a new epic project since the car!

I love seeing you learn. I appreciate the way you edit your videos to give us this perspective. "It's just a few lines of code..."

He’s back!

Finally!

Epic

Thank YOU I’ve been waiting to see someone do this since the keynote!!!

Its been a while but it was worth it!

you can tell a computer is driving just by the wheel turning movement

Bro this looks cool Imma try too