Re:Webcams aren't good enough
by adhesive_wombat on Sunday June 19, 2022 @05:59PM
(#31803946)
Maybe in 2002, but it's a USB webcam in 2022, if you're writing your own drivers from scratch on any platform rather than using the standards-compliant ones, I'd first wonder if you haven't got a case of NIH syndrome rather than an actual case of being constrained by the standards.
But maybe they genuinely cannot use standard protocols for some reason, then sure, writing a Linux driver might not be worth it.
I'm half assuming they do something special, half responding to the comment which describes them as a hardware company. But yes, ideally a webcam should have the basic functionality available just by plugging it in, no special software needed.
Re:Webcams aren't good enough
by GekkePrutser on Sunday June 19, 2022 @07:45PM
(#31804661)
The postprocessing should be done in the cam though, IMO.. And it should just give the result back ideally in a standard format.
But I would prefer better optics and sensors over AI stuff like many others have mentioned. When you have good video you have much less need to fiddle with it.
I worked for a company that did pre-infiniband HPC interconnect drivers (Myricom) in the 2000s. We supported N-different *nixes (Linux, Solaris, FreeBSD, Tru64, Aix, and probably more I'm forgetting), Mac OS X and Windows in the same codebase. At least 50% of the driver and 95% of the userspace code was OS independent. In order to do this, our drivers were not "typical" drivers. Eg, we used ioctls everywhere (even on OSX) for driver/library communication.
The software postprocessing is a different question and could be more common... But you're still specialising it for DX / v4l2+gui / whatever macos uses. That said... I'd kill for an independent library for video processing that specialises in camera outputs - beyond different profiles, there's no technical reason that couldn't be used independently of hardware.