This is very interesting! This detector promise you " no time / filament wasted " in print failures. Check it out.
I don’t like the cloud aspect of this.
Yeah, I know you can setup your own server, but why can´t the Pi, where OctoPrint is running on, do the same.
Speed wise this shouldn’t be an issue. I takes a picture and even so it would take a minute to analyze one image and you need e.g. around 10 positive pictures to avoid false positive, you are around 10-15 minutes until detecting an error, which is far enough.
Especially a Pi4 should have enough cores and power to handle the detection itself and serving OctoPrint. Having a server is just adding so much hassle to it.
Unless they add(ed) a “run here”.checkbox to the settings since I took a look last time, then I find it currently to complicated to setup.
I think part of the logic behind the cloud connection is that it is using input from thousands of other users for analysis by the AI engine. It’s a “crowdsourced” effort that saves you having to create images of thousands of successes and failures to train the engine on your own system. As the training data is refined, it may be possible to download just the data, but I’m not aware of a lot of efforts to do this other than Spaghetti Detector.
For me, at the moment at least, having a camera on each of my 2 printers is enough. I can check in on them from time to time without much issue. I can see, however, how this might be very valuable if I ran a printer farm. Having dozens, or even more, printers be able to shut down automatically on print failures would be very attractive.