HOW does person detection actually work?

I’ll preface my question by saying I have super weak internet, 6/1, and it’s been extremely laggy, seemingly only after I purchased cam plus for person detection…

I can’t find a definitive answer on how exactly person detection works, I only have V2 and V3 cameras so no on-board detection and I’m mainly talking outdoor cameras - which would constantly trigger from snow and such if I didn’t have vehicle and person detection enabled. So, are the cameras constantly uploading video/images to the cloud, determining it’s not a person or vehicle and not sending me notifications? If yes, could having a dozen or so cameras doing this constantly be the reason my internet has been laggy?


Since you just purchased Cam Plus, I would say yes 12 cameras could be stressing your network. Each camera could be uploading long videos with no cool down time. Here another thread on data usage.


Based on the Detection Sensitivity you set in the cam and your use of Detection Zones, when the cam is triggered by any motion on CamPlus, every motion video is uploaded to the servers regardless of what is in the frame or what miniscule motion caused it to trigger. The video is uploaded so long as the motion is still happening. These videos are full length up to 5 min in length.

The number of activations and the length of the videos can be determined by looking at the Events Tab without any filters.

What AI detections you have set, or what AI videos you have set to save on the server, has no affect on the number of uploads. All motion videos are uploaded. Based on your settings, the server then AI interrogates the video for AI tags, notifies you, and saves or discards it.


My Wyze Cam V3 had this event recording today, which I found very interesting. If the video didn’t upload, how did it know a person was detected?

I was running a test to see how long it would take for a rule to trigger the floodlight to turn on after person detection, and it was quite fast (<10s). But after my test when I checked the event recording it was only a still shot, with this message when I clicked on the help icon on the bottom right.

I think I read somewhere that Cam Plus would start AI detection before the full video is uploaded, while Cam Plus Lite would wait until the full 12s video is uploaded. So perhaps a thumbnail is sent right away, then the full video for full analysis?

1 Like

It doesn’t. Mine is constantly mistaking vehicles and palm fronds as "people ". Its pretty useless

I believe this is the case. Cam plus allows you to view in progress events and they have the “live” tag, but cam plus lite just throws an error or doesn’t show the event at all when it’s not done uploading. I also believe the AI interrogates the initial frame (thumbnail), and also later in the video. So that could be why you are getting AI tags on a failed to upload event.

This is a failed upload message due to an error that happened during the video upload process.

The AI bot does begin interrogation when the upload starts and will tag and send the notification immediately upon verifying the object as AI, even if the video has yet to finish recording and uploading.

This is an event video upload that started, was tagged, but then failed to finish uploading and saving.

If this repeats or becomes a pattern, logs will need submitted so the Devs can look into the cause.

Thanks @IEatBeans and @SlabSlayer. This is helpful to know.

This is the first time I’ve seen this upload issue on my Wyze Cam V3, although I’ve seen quite a few on my Wyze Cam V2s (on Cam Plus Lite) until the recent 1851 firmware update. Hopefully everything will be stable and improve after the maintenance update…

1 Like

You’re welcome.

I had experienced those ocasionaly in the past on my old original router (2 upgrades ago) when I was streaming very heavily and had a barrage of multicam events all at the same time. It was just too much for my router and uploads suffered.

Actually I did see this happen on my OG (worst signal among all) the other day, but when I try to open the event recording I get an error (even though it’s on Cam Plus). It seems the signal was so bad that the Cam couldn’t even load the initial thumbnail. The V2s beat the OG in terms of timing and successfully uploaded the videos, but not the OG.

1 Like

Well that’s unfortunate and likely the culprit then, even at lowest sensitivity blowing snow, dust or bugs constantly trigger them, so the camera(s) will be uploading pretty much 24/7 - Just not notifying me.

Time to upgrade to all V3 Pros I suppose.

1 Like

How is that going to improve anything? Even though the V3Pro has Edge AI on board, it operates in exactly the same manner. All motion videos are uploaded. The Edge AI is just a “pre-filter” that helps the Cloud AI Bot to tag faster and more accurately.

1 Like

If you have a V3 pro with no subscription (I.e. no cloud AI), it would probably generate more false positives than a V2 or V3 with Cam Plus Lite and cloud AI person detection.

1 Like

It certainly won’t be as accurate without CamPlus. But I can’t say with any degree of certainty that it would be more false positives. It may be more since the Edge AI isn’t as robust. Then again, you may also have more missed positives with the less powerful AI. :man_shrugging:

I was focusing more on the number of uploads. If the Event Recording is on, every event uploads. Without CamPlus, there will be less uploading because of the 5m cooldown lockout between uploads and the upload of a thumbnail snapshot instead of a full length video, so less bandwith used, but that is a really big tradeoff.

1 Like

And as long as you have an SD card in the camera you can go back and see the full video of what happened some people don’t believe in an SD card and they can’t see it