Does this AI even get better with all the video submissions?!

Who else is having the same issues with capturing every bit of movement that isn’t wanted? I.E. Waving trees, light changes, trees sounding like they’re “talking”, etc… If there’s an AI department, they sure are slacking off a ton…

I can confirm that it definitely does help.

For example, I had 2 cameras that used to ALWAYS detect my black cat as a person at night. I submitted corrections for these and over time it went from EVERY single time tagging him as a person, to less and less often until finally they stopped detecting him as a person. It didn’t fix it all at once, but I absolutely saw it change from 100% of the time to nearly never over the course of many months.

You have to keep in mind that with the way Machine Learning AI works, it’s already had millions of other events put in to it’s training, so a few submissions aren’t going to immediately outweigh the millions of others it was already based on. It will take time and consistency and similar submissions from others to improve similar situations.

And yes, they do have a dedicated AI team, or at least they used to. They even gave us an AMA event with the AI team. Who knows how often they change or reorganize things, but they definitely did/do have an AI team, and I’m sure they still do since they told us that this year would have a strong focus on AI.

For what it’s worth, they only update the AI training data once per month.

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.