Summer is ending and we hope you all had a nice time in September. We are entering the rainy season in Seattle and still working from home like we were last year. A recent peer-reviewed study found that the pandemic has extended the average American knowledge worker’s workweek by 10%. How’s the situation in your area? Hope everything is going well and you’re keeping safe.
We continued the model training work this month. We are seeing major improvement for pet detection for medium-sized pets together with improvements to the person, vehicle, and package models.
For further model improvement, we will look into several specific fail cases and work on plans to improve the success rate for them.
This month the new system was rolled out to thousands of devices for a larger test group. Based on the feedback we’ve collected so far, it’s 85% performing better or no worse than the existing AI system. We’ll keep closely monitoring the metrics and decide if we’ll roll out to other devices in addition to Wyze Cam v3 for a larger-scale test.
Thanks for all your feedback for Face Recognition! Several issues have been identified and will be fixed in the next app release (app version 2.25). At the same time, we are working on improving the success rate for detecting and recognizing the faces from video (especially for doorbells).
Thanks for all your feedback on last month’s discussion. We’ve included the feedback into the model training and the testing is still ongoing. You can find the sign-up sheet for sound detection here.
We’re still looking for the data examples below as we build sound detection for Wyze parents! If you have any examples of the following videos, please send them to us at your convenience to help us train the model.
- Videos with babies inside the frame
- Videos with audio of a dog barking
- Videos with audio of a baby crying
You can submit the video using the banner at the bottom of your Event video.
Shawn from the Wyze AI team