Telegram worries by Jan-Lukas ElseJan-Lukas Else (jlelse.blog)
Telegram was always my favorite messenger, as it provided a fast and user-friendly experience on multiple devices. But one and a half years ago, I already started doubting my decision to rely on Telegram. Now I question it even more. With the latest release, Telegram introduces an in-app browser, wh...

The direction Telegram is going is unsettling to me too. I always preferred this over any others because it was simple and just worked™ on all of my devices. Including my Linux PC. I wish I could name an alternative but what? The usual suspects, Threema and Signal suffer from the same problem: Centralized services.

And Matrix? OMG. I wish I could recommend that but the on-boarding experience – even for tech savy people working in this industry – is the worst. And sadly that’s not even a question of the client used. Each requires a high suffering tolerance from their users. It feels like it has all the right ideas in the most complicated way possible and while I didn’t host any Matrix serves myself I hear they have a very heavy footprint.

I’m flabbergasted how good this tracker-neuralnet plugin for works. It does the with just a webcam without any clips, reflectors or LED stripes. I kinda expected this to not work really well in a dark room, that I prefer for gaming, but I was wrong. Even with a tiny light in one corner of the room only it kept tracking flawless.

…can even scratch my nose and it keeps tracking.

To get this neuralnet tracker input in the first place I had to download the ONNX runtime package onnxruntime-linux-x64-1.18.1.tgz from https://github.com/microsoft/onnxruntime/releases/tag/v1.18.1 (My Fedora offered 1.15.1 from it’s repo but this was at the time of writing not sufficient and having it installed resulted in failure due to an API mismatch). I didn’t even install it somewhere, just unpacked it in my Downloads folder.

Back in my OpenTracks folder I run cmake the way I’ve done it before several times but this time I added the unpacked onnxruntime folder to the config.

Configure did it’s magic (note how it picked up module tracker-neuralnet once the ONNXRuntime_DIR was set) and here we are one make later. This is rather impressive 🤓

YMMV