I haven't posted for like... 1.5 years, so it's high time I write about something useful or entertaining. Well, this post is not going to be that, and instead I will be talking about an obscure piece of code written in a dead language.
We've found a drop-in replacement for the Nreal Light, called the Grawoow G530 (or Metavision M53, and who knows how many other names), so we finally have 3 more protocols to write about in this blog.
I felt a need to apologize about the site looking amateurish in a client call for the second time in a week, and it was kind of a red flag. We don't want to have a bland, cookie cutter 2020s company info page, but we do need to make it less outdated, while keeping the ａｅｓｔｈｅｔｉｃｓ.
It was time to integrate support for the Nreal Light's stereo camera into our Android build, but our go-to library (nokhwa) does not work on Android. In fact, nothing seemed to support this specific camera on Android... So it was time to make our own USB Video Device Class implementation (but not really).
The whole FIR thing has served us well for the last few months, but some new developments has forced us to look into alternative prediction algorithms. Thankfully (?) I'm on many Discord channels, and someone mentioned Double Exponential Smoothing on one of them. Everyone loves a good clickbait, so you won't believe what happened in the end...
We are experimenting with various VSLAM algorithms, and a big part
of those is always "bundle optimization". While it is just spicy
Least Squares with sparse matrices, I couldn't find a well-maintained Rust
library to do this. I, however, found g2o, a c++ library which is both
very well structured, easy to use, and flexible. The only problem is
that it's not Rust, it's C++...
I've just finished implementing the third completely different USB protocol for my open-source AR glasses drivers project, so it's time to summarize my findings, and drop some fun facts on you in the process.
It's been a long time since I've been to any security CTFs, even longer time since I've written my last writeup. But, as life would have it, I came across a task during my AR adventures that was exactly like a CTF task:
I've found a non-standard CRC checksum on a firmware file, and had to reverse-engineer its parameters.
One of the less fun parts of trying to run a company is having to create all kinds of marketing materials, such as product and investor slides.
I already use reveal.js for all my slideshow needs, so there's one thing that will make the job bearable:
A scrolling synthwave-style background made with HTML+CSS.
I always wanted to do a rally-style demo for the Augmented Reality glasses, but first I'll need actual pace notes to display.
And there's no way I'm doing that manually.
Still field testing, still not satisfied with the image shake. IMU prediction helps a lot, but it is very inconsistent: sometimes it is rock stable, sometimes it is lagging quite a bit (in the delay sense).
I've also noticed the definite lack of tearing, even though I'm using Immediate presentation mode, which should be low latency, but should also cause tearing.
The lack of tearing means I still have unnecessary latency...
I've noticed a weird distortion during field-testing the new graphics on the AR glasses: signs that should have been far away looked way too close, but they got further as I approached them. Pupil distance didn't seem to fix it without introducing other issues.
So I've decided to finally properly calibrate the projection parameters.
Content warning:
All drawings are really badly made
with KolourPaint (KDE's version of MSPAINT).
The main problem I have with the current AR glasses is that it has a motion-to-photon latency of around 35ms. I don't think this can be any lower, as this is 2 frames, and I'm pretty sure the GPU->DP->Oled_Driver->OLED chain takes this long.
So it's time to finally implement proper lag compensation in software.
You can probably see that the logo at the top left is animated. At least if your browser decided to cooperate. This was done by using a .webm (or .mp4) video instead of just a regular image.
But how was the logo made? The answer is not "with an image/video editing software a normal person would use"