AI modeling is designed to be a constantly iterative learning process. A month between releases is absurd. Your backend systems need to be updated in real time. One of my cameras no longer picks up people and vehicles—and that’s pointed at my front driveway and yard. Which, what’s the point? Another picks up insects and is always notifying me. I’d rather see who’s approaching my house than count the number of bugs that buzz by my camera. I consistently upload my videos to Wyze with what these cameras should and should not be recording and notifying me of. Which goes back to my original point. Everything submitted to your modeling algorithm should be instantaneously become part of your learning model, parsed and deployed. I’ve already disabled the bug-counting camera and no longer depend on my vehicle/person camera. Again, what’s the point of having these products? Rather than pushing new offerings out to completely differentiated market segments seemingly every week, why not perfect what you already offer and then move on to your next initiative? This entire Wyze experience (including not just my cameras, but also two now-mothballed Wyze “smart plugs”) has become an exercise in futility and frustration. I’m actually rooting for your success, but please do better.
I completely agree! The AI is basically a joke. It constantly does not see pets and often people, yet see’s parked vehicles. Seeing a parked vehicle is a useless “feature”. Wyze really needs to stop adding things and fix what the already have, that doesn’t work correctly. And…stop using AWS because these outages are really getting old fast!
AWS is one of the biggest scams in who knows how long. It’s inherently buggy and lacks any genuine security safeguards. But these screw ups are ultimately the fault of Wyze and its AI contractor. Seems they have little knowledge about how machine learning should work. Data input, data interpretation, data parsing, data output and subsequent reaction based on data. That’s the simplified version, but this type of process should be completed in near-real time. That’s how most companies leverage AI. Banks, for example, need sub-second AI input and feedback to monitor both existing fraud and new forms of fraud. Otherwise, ¯_(ツ)_/¯, expect to lose money. Wyze and contractor may not have the resources to support the hardware and software needed for this type of data-feedback system loop, but they should look at this as a long-term investment and commitment that benefits their entire product portfolio. Otherwise, people like you and I are going to abandon the platform. So, to Wyze, invest now and benefit later. After all, it takes money to earn money.
It’s been how long now and the Wyze AI is still only about 30-40% correct. I have my vehicle parked in the street and my camera watching it. When other vehicles go by, the motion is detected but the video isn’t labeled as anything. But when people walking their dogs go by, the video is labeled as a Vehicle and sometimes it also adds a Person but never see’s the dog. It’s so hit and miss that I just don’t understand why Wyze is adding new label detections when the current ones aren’t consistent enough yet. I’d thing if they can get those accurate, then you add others. All I’ve wanted was for my motion detection to notify me when people are near my property, that’s it but it doesn’t always see them so I end up with a lot of video to sort through. I might as well just let it record any motion and turn off all the labeling junk.
Good Morning - l have 6 cams monitoring the activity around my yard, mostly for the wildlife we have in rural NH. I consistently submit videos to help in tweaking the AI. There are many times that animals and things (from raccoons to insects in spider webs) are tagged as “Person”. I just untag those and submit. Many times it thinks wildlife are pets (raccoon, possum, fox) and I do the same. May I suggest another category for animals or wildlife? Birds, bats and owls could be tagged as well?
Thank you for your service - Paul
Same. A couple weeks ago, my PanCam v1 with CamPlus (pointed at driveway and street) straight up stopped doing any AI detection at all. No changes to the settings; everything is still enabled correctly, but no vehicle detection, person detection, package detection, or pet detection. I’ve disabled and re-enabled CamPlus. I’ve restarted the camera. I’ve turned off and on detection and notification. Still broken. If it does not magically turn on again in a week or so, I guess I will have to cancel CamPlus.
You are not alone with your problems w\ the WCPv1.
I was working w\ Wyze in the DM to find the cause to some CP\CPL recording issues when my tests revealed that the cam is also not activating on motion as it should or tagging AI as it should. Same as you have reported.
The issues were never resolved. And I tried every remedy you did and then some. I am going to delete and reinstall when I get time, then flash back the FW to see if that fixes it.
Here is the original thread:
Thanks for the confirmation. Appalling. I am glad you still have the patience to work with Wyze to resolve this, but I am afraid I do not. My CamPlus subscription was going to renew next month, but I am canceling it since CamPlus Lite does a much better job for less money.
Sorry it isn’t working out for you. I only have the one pan, the rest of my many are V3 and one VDBv1. All of the V3 are working as expected and have been improving. I have the CP unlimited license, so one cam acting up until they fix it is inconvenient, but not the end of the world. The pan is indoors and very low use also as an intruder alert cam only when I am away. Will post back when the status changes
Same. One Pan v1, one VDB v1, one GDC, and four Cam v3. I’m not inclined to pay for Unlimited when the one I do pay for does not work. Since CP Lite works well enough on the other cams, I may as well stick with that. Cheers!
Don’t need more categories that won’t work!
@skst, I have posted this issue with the WCPv1 short recordings even though it is assigned to CamPlus to this month’s Fix-it Friday thread.
If you are still experiencing this issue, feel free to follow the link, like the post with a heart coin, and add your experience in a reply post (if you are so inclined).
One of the labels is for insects, yet none of my camera’s has ever labeled them, yet they get recorded. What’s the point of having the label for insects if it never labels them?
Today I had 30 recorded events that were all due to insects, yet none were labeled as insects. I’ve been submitting to Wyze for a year now to what was recorded but the AI still isn’t learning.
It’s disappointing to hear that Wyze AI is still not learning as expected. Continuous improvement and learning are vital for any AI system to deliver optimal results. Hopefully, the team behind Wyze AI is actively working on addressing this issue. In the meantime, it’s crucial to provide feedback and stay updated on any developments. Let’s hope for a positive turnaround and enhanced learning capabilities soon.
Honestly, I have no confidence that Wyze will fix the AI, considering it’s been several years and it’s still not fixed. Again today my V3 camera was motion triggered by a person walking their dog, but the recording was labeled as a vehicle because a truck was parked in the street. And then when the truck left, the camera didn’t record it leaving even though the entire truck was within my motion grid. This is so ridiculous! The team at Wyze that is working on the AI has no idea what they are doing!
Exactly! What’s the purpose of all the other ones? I send what in the video faithfully. I’ve never seen anything label insect, waving tree etc.
There are currently no labels (tags) on event videos for the others yet. By submitting your tagged events for these others, you are helping build the AI model for possible future implementation.
Additionally, if you have a video originally tagged Person, but it was actually a spider, changing the tag from Person to Insect allows the AI Model to be retrained so that insects aren’t identified as Person in the future.
I still get person cs spider after many corrections submitted.