Some of your posts are getting held for moderation because you're a new user, and I'm approving them in between doing other things. I found the message that was missing.
Thanks for the report of the dog which was really a squirrel. I've submitted that as an example to add to the training.
I've shared the two cameras with you that generate the alerts, along with what I think are enough permissions for you to be able to see the green box. I have not added permissions for you to be able to rate the alerts (which would be required for you to be able to submit wrong examples yourself) because I don't want you to inadvertently rate alerts and potentially change the behavior of the demo. For any wrong alerts you see please let me know, either here or at
sup...@camect.com if you just want to send a list of examples.
As far as picking cameras, we were limited to what we could find publicly streaming live on youtube or elsewhere. Most cameras streaming live to youtube are in busy touristy areas where we can't do an interesting demo because all you'll get are a constant stream of person alerts. The other problem is that many are movable cameras, and sometimes they're also often much farther away from the subjects than in a typical home. (e.g. The top of a tall building.)
The "AXIS 4K" one seemed reasonably close to the scenery at many suburban homes, and in particular has plenty of wind, rain, and bugs to show we don't misdetect on those.
The golf course was just another location where we could show some vehicles, with objects of interest not overly far away, and show that the statue is not misdetected as a person. It used to have a lot more activity before Covid-19 happened.
If you're aware of any publicly accessible cameras that approximate the scenery of cameras at your own location, we're happy to try adding one and see it will help whatever issues you have been seeing. We could draw exclusion areas as needed, but we wouldn't put it into the demo if we do that.