AMA SUMMARY FOR: "We're the Wyze AI team - ask us anything (again)! - 7/14/22"

(Thanks to the AI Team for taking the time to give us all these updates and answers!)

We’re the Wyze AI team - ask us anything (again)! - 7/14/22

AI recognition of Moving Vehicles (not parked vehicles) and lights/shadows triggering motion detection

Ignore stationary vehicles

ETA on alerts for only moving vehicles (not stationary ones)?

Plans to speed up notification arrival?

You used to have an Edge AI job position posted, are you now actively developing local/edge AI solutions?

Are there plans for onboard AI for future Wyze Cams? That would speed up the notifications and save Wyze money without using the cloud for events.

Will you work on processing AI detections on a local device to get notifications faster?

Instead of motion in the detection zone causing the AI to detect the entire screen, how is progress on having the AI only scan detection zone areas of the screen too?

[So to help forum users know why something wasn’t detected, we may now we need to ask to see their detection zones, since that could be the reason an object in view was excluded from detections.]

Tell us about the feature to show us what the AI detected and tagged as an object?

Another community member who couldn’t attend this wanted me to ask for him: “Does the [event submission] feedback actually help when submitting the correct feedback [or detections]? I have spider webs that will be labeled as people all the time and submitting the correct feedback for the last 8 months or so has not really improved much.”

Tree limb shadows being detected as a vehicle in the backyard

Floodlights triggering motion alerts

What are the chances of getting an upgrade to the detection zone (for the AI, motion, etc)? This second version of the D.Z. being a grid is a HUGE improvement over the original singular box detection zone, but it is still not as precise as many competitors’ methods allow and as many of us would really prefer. We can’t do precise lines at angles, etc. This could also be more important if you come out with higher resolution cameras as your VP mentioned in his AMA that you would have some cameras in the future with higher resolution (in that case, each D.Z. square would potentially be covering twice as many or even more pixels now).

Ability to add or remove multiple Faces at a time (instead of individually).

When is friendly faces coming to the doorbell?

Will the Matter Initiative have any influence over how you handle AI or anything else related to your team?

Did you finish “Optimizing the pipeline” to allow the AI to analyze the entire event for all faces and detections now (not just the first one)?

Is your AI using neural networks or rules based logic? And Can you give clarity into what are some hurdles you need to overcome in order to reach some of your milestones for the coming year?

Wyze Anything Recognition / Smart Vision updates?

Current Face recognition limits are 300 pixels to maintain a 90% confidence interval = about 6ft away. How will you expand and improve these limits (higher resolution, or other new innovations that Steve McIrvin mentioned)?

What other AI detections are you considering having as global options (trained for everyone to have)?

I would love more animal categories (ie: predators or individual species) instead of just everything being “pets”

Current Doorbell concerns: Person notification delay, Notifications outside of detection zone. iOS thumbnail previews stopped working.

Why can’t we edit subscriptions for Cam Plus purchased through the web and the app all in the same place?


Many thanks for doing this carverofchoice! As always… well done! thumbsup