AI Vectorizer : Server is probably booting, please give it ~10 seconds!

[stefankinell]

Hi there,
Nice idea. But it does not work very well. Please update this error message. I have been waiting 10 minutes. And then ten minutes again. I get no response. I have tried multiple times now.

2024-10-03T09:46:09     INFO    Bunting Labs AI Vectorizer : Settings saved successfully.
2024-10-03T09:46:12     INFO    AI Vectorizer : Server is probably booting, please give it ~10 seconds!
2024-10-03T09:46:14     INFO    Arbetsuppgiften misslyckades : AI Vectorizer processing map chunks on server
2024-10-03T09:46:17     INFO    AI Vectorizer : Server is probably booting, please give it ~10 seconds!
2024-10-03T09:46:17     INFO    AI Vectorizer : Server is probably booting, please give it ~10 seconds!
2024-10-03T09:46:22     INFO    AI Vectorizer : Server is probably booting, please give it ~10 seconds!
2024-10-03T09:46:27     INFO    AI Vectorizer : Server is probably booting, please give it ~10 seconds!

[brendanashworth]

Sorry about this — looking into this error now.

[stefankinell]

So I managed to get past the booting, after having waited 10 x 6000 seconds ;-)

But it is really really slow. I can trace like 20-250 meters, then I have to wait 20 seconds for another square to load. The tool looks nice, but it is not four times faster than manual digitizing this way. The other way around for me. I might be a quick and experienced mapper and your claim to do it four times faster might be for a beginner that never have drawn a line? Sorry if this sounds sarcastic,I am just wondering if there are still bugs, if the server is not working for you tonight or if this is what can be expected. I was trying this out on the demo. It is a rather expensive tool, especially when it does not work.

The other concern I have is that the lines that I get are rather busy. Below is a trace I did of a more or less straight line. As you can see it is jumpy and is sort of plaing pin ball between the outer borders of the line I was tracing. Again - is this to be expected? If so I have to do quite a lot of post processing on the traced lines to get it in working conditions.

[brendanashworth]

Sorry it's pretty slow for you. For reference, inference takes about 660ms on a new map chunk (which is then cached) and it's another ~200ms for other computation, so there's definitely some bug here. Would you mind emailing your API key to brendan at buntinglabs dot com, so I can take a look at our internal telemetry and see if there's something wrong?

If the line is that busy, you need to zoom out when you digitize. Performance is very sensitive to your zoom level. This tutorial video contains some important information: https://www.youtube.com/watch?v=PKEuQS4sMJE Unfortunately we're not sure how to make it work well regardless of one's chosen zoom level yet.