Wildlife detection for Cam Plus

Add wildlife detection to cam plus. We often have deer, raccoons and occasional bobcats wandering around at night that we like to keep an eye on.
Another option might be to expand the pet category to animals in general, since the ai might have a hard time distinguishing a cat from a raccoon (or a pug from a pig or loaf of bread ;))

Welcome to the Wyze community @zedn4c3r!
I believe pet detection is currently in testing, but detection of larger wildlife and specific animal detection would be great!


Add AI events to detect wildlife - either generic or specific, such as deer, rabbit, etc. This way a camera can be used to trigger wildlife deterrent actions, like turning on sprinklers, or playing a dog barking sound on an outdoor speaker (or on the camera itself).

1 Like

Pet vs. Animal or Critter

Could “Pet” be universally changed to “Animal” or “Critter”? Would everyone who monitors their pets be irritated to have them called something else? It. sure would solve the massive. amounts of wildlife being improperly categorized. Do it once through software and be done with it.


Ideally wild animals would be a separate category from pets, though it may be hard to distinguish them. But that would be great for home automation if you have pets and wildlife is visiting your backyard. I would take separate categories even if the distinction between them is not perfect. I would also be happy if there was a single category, but it was able to detect rabbits. Rabbits don’t come up at all in AI detection (for me).

1 Like

I agree - ideally. But for a quick fix, I could use just one critter category instead of calling my humming birds and possum and raccoons and fox as pets. Much more difficult to do through AI. How could it ever tell the difference between a coyote and a house dog?

1 Like

How about renaming “PETS” to “PET/ANIMAL”?


I wish that in addition to giving feedback to wyze with tags so it can create its own recognition system there was a way for me to add personal write-in tags and then quickly filter and search videos in my own event library. Basically just add a tag to create my own filter category after I watch.

I am an avid birder and trained ornithologist, which means I’m an expert at even hard-to-identify species. When I go through my videos I’d love the option to save them under the species I identified so I can go back later and search for “bird” instead of “chipmunk” but also more importantly filter things that I wouldn’t expect even the best recognition software to know, like specific species or subspecies of animals. Or even specific behaviors of interest I’ve captured (for example: foraging, feeding young, bathing, aggression displays, courtship, etc.) This would be pretty easy to do I think and wouldn’t require adjustments to the recognition technology itself at all.

I use wyze outdoor cam often to teach about species identification and bird behavior, and saving all the videos and then sorting them myself can be a pain when I feel that it could be an easily built in feature. Does anyone know of a way to do this in wyze already that I haven’t discovered, or if it’s something wyze might plan to add one day? Even Apps and cameras that specialize in species ID do not have the accuracy of an expert so I’m not looking for that. I think it could just be a good feature in general if you are using wyze for fun way to capture moments (more than for security, as it was originally designed) as I’ve learned many people use wyze for recreation as well.

I guess this might be a different wishlist request entirely, just the addition of adding personalized filter categories to your vids.


AI Request - Wildlife Event Tag

Deer and birds in my yard are detected as pets. Would love to have a wildlife detection option. That way maybe a rule could trigger the siren when they start eating out roses!

[Mod Note]: Your topic was merged to this #wishlist topic for better visibility. An AI model for wildlife detection needs to be developed to give meaning to a wildlife event tag and associated rules. Please remember to scroll up to the top and click the VOTE button to show your support.


In this month’s email newsletter, they mention a “3D printed a Wyze birdhouse.” It seems with the success of competitors, this would be an obvious thing for Wyze to productize:

  • AI bird/wildlife detection for the cameras
  • Bird houses, bird feeders, nectar feeders, with mounts/magnets for the weather resistant Wyze cameras. (Outdoor cams, V3, OG, etc.)
1 Like

Identifying wildlife

Living by the foothills, I get a lot of wildlife visitors including skunks, opossums, coyotes, bears, and even stray cats. What’s confusing is some of these animals are identified as a person, primarily the skunks and opossums or they are identified as pets like the coyotes or stray cats. Is there a way to have some new designations?

Wildlife for the skunks, opossums, coyotes? This could even cover raccoons who also come around.

Bear, specifically for bears

Cat for stray cats or cats who aren’t neccesarily your pet cat

Dog for stray dogs or dogs who aren’t your pet

To be honest, seeing a motion designated as a person or pet at 2 am outside is really alarming, only for it to be a false alarm.

Does anyone else have these same problems?

[Mod Note]: Your topic was merged to this Wishlist request for consistency in grouping similar requests. Please remember to scroll up to the top and click the VOTE button to show your support. You may also wish to comment on and/or vote for Animal recognition and tracking.

1 Like

So in our area, there are bears and cougars, and sometimes they wander around the day, which is a safety issue, especially for smaller humans. So if Wyze could add this feature, I’d very much appreciate it. This is one reason why I installed the cameras, and now I’m using various workarounds, such as movement detection, but it does not work properly. Please, please, implement this ASAP.

1 Like

THIS. It is really really hard to categorize animals by species – even for people sometimes! I cannot imagine Wyze Smart Filter being able to do this.

Also, some of backyard categories are already filtered as pets automatically. In addition, people have companion animals that are similar to or nearly the same as wildlife (birds for example).

Having a blanket category for “Animals” or “Critters” seems like a really smart and easy solution to this problem. It would really just be a matter of dividing moving living things into human and non-human. And besides the fact that Wyze often tells me that my Red Bellied Woodpecker is a person, Wyze is already pretty good at recognizing humans, and many pets, so it seems like a realistic filter idea.

…btw I’m inclined to agree, woodpeckers are people too :laughing:)

EDIT: Sorry - I meant to reply only to PaulT’s comment of just changing the “Pet” category to “Animal”/“Critter”. I think that asking for an AI system that essentially does what iNaturalist tries to do is sort of absurd. Everyone who has ever tried to identify a plant or animal with an AI app (Merlin bird ID, iOS look up, etc) knows that there are tons of problems with that, and you are basically asking Wyze to do something that even expert biologists sometimes struggle with.

1 Like

Thanx Carly - I found a solution: ignore improving the AI altogether. I stopped playing the game a while ago.

1 Like

Anybody looking for near term and realistic solution for this … just don’t lol. Unless you have time to learn how to train your own model (actually pretty doable in 2024, if you have $2k (ish) to spend and a lot of time). Even then, object detection at this level of specificity, on the best models are probably 24 months away.

So, this is not a Wyze issue, and I’m shocked they’ve kept it open, as I always understood “maybe later” to mean “outside of our 6-month CI/CD pipeline” (that’s a fancy way to say we’re already booked up for any development for the next 12 months).

@WyzeTeam take this down from “Wishlist” – it’s nearly impossible to disappoint people more than you already do, but this is one case where disappointing people literally isn’t your fault.

Naw, there are actually several totally free docker containers you can use to do your own detections from camera streams. Lots of them will even run on something as affordable as a Raspberry Pi or any old computer. Then use something like Scrypted or Docker Wyze Bridge or another similar container to grab the RTSP stream and feed it into one of the free image detection dockers. Lots of people do this for Home Assistant to have their own detections. Some of them have a ton of animal detections already built in. For example, if you run Docker Wyze Bridge into Frigate, you can have the following animals already detected:

  • bird
  • cat
  • dog
  • horse
  • sheep
  • cow
  • elephant
  • bear
  • zebra
  • giraffe

and many, many more detections…all free. And if you just run it on an old computer, no hardware cost either, otherwise EASILY just buy an old R-Pi or similar for under $50.

Anybody could do all that right now with their webcams if they don’t want to wait for Wyze to implement something.

I’d say cost is nothing to max of $50, and time will really just be a few hours or so. give or take a few hours depending on your level of understanding or experience with docker containers.

Yeah, I have that, all of that. Docker-wyze-bridge → frigate → scrypted → HKSV. With some 4k Amcrest cams as well. I have it running on a Mini-PC with a PCIe M.2 Coral TPU – literally a dedicated chip for AI/ML applications. That system has seen thousands of animals, and please trust me when I tell you, that we are nowhere close to being able to distinguish wildlife from pets.

It can tell IF there is an animal, about 90% of the time, but every single day have at least 1 bird on my birdbath cam get identified as a person. That’s a weird scale and the models weren’t trained to detect birds that take up more than half the view frame, so it’s understandable.

I put a camera in front of my TV and it watched animals (National Geographic, PBS-Nature, etc) for 72 straight hours. It saw all of those animals (the list is actually longer than that), many times. And everything is a horse lol, unless it’s actually a horse, which is often a deer or a dog.

They get better every day, and if you trained your own model, it could learn to pick out and differentiate YOUR animals in a week. But telling wildlife from pets? We are not even close.

1 Like

@carverofchoice Got this gem tonight and it reminded me of this thread. This is my wife taking out the trash. This camera is actually running on an M2 Apple Silicon processor with a 32 core gpu and a 16 core neural engine (TPU), running on CoreML framework. This is as good as it gets in 2024. And this is a terrible example of it’s abilities, but also, kind of a great one lol.

Never seen it miss like this with a human, and it’s obviously because of the bag, that is a HIGH confidence detection that my 125lb wife is a cow. Actually, two detections.

Again, not a normal example, never seen it above 50% confidence with an error this bad; but, getting back to what this thread was about – the best AI on the best consumer hardware available has never, ever been able to tell a dog from a coyote.

And given that many humans commonly make that mistake, will we EVER be able to confidently and reliably tell a dog from I coyote from a black and white 2k video? I’m not sure… In a high-res color still photo, sure, we’ll get there someday. But not on a back and white 1080p cloud cam.

Never thought I’d be standing up for Wyze like this, but it’s not their fault. Where the wyze team has messed up, is by marking this as “maybe later”.

And I can’t tag at-wyze-team cause they blocked me. Classy move. at-team-wyze I’m actually defending you here lol

1 Like

OMG! :rofl: I was dying laughing so hard when I saw that label on the detection screenshot. I don’t think I would dare show my wife. Thanks for sharing.

I haven’t used it yet myself, though I plan to once I catch up on work and setup my new server. But can you set a limited confidence interval? From things employees have said, I am assuming Wyze likes to keep their AI detections closer to a 90% confidence interval for at least some detections. At least they talked about how face detection needs 300 pixels for a 90% confidence interval, which ends up being like 3 meters away on average. Are you able to set a minimum confidence interval for your detections?

Interesting. I’m assuming they just don’t want people constantly spamming all the employees like has happened in the distant past. They’d rather an individual be notified of something important. I didn’t realize this was ever restricted.

1 Like

lol my wife has a pretty great sense of humor and enjoys watching my tech failures :wink:

So, with scrypted, you go far, far beyond confidence interval. You can bypass the camera’s motion sensor altogether (with wyze specifically, you actually have toemphasized text**) and it’s a whole new level. You have motion detection settings like Wyze has in their backend development, but YOU get to fine tune it perfectly for you. There are unlimited zones. And they can all be set differently in one camera:

My bird feeders zone has settings that literally just detect bird sized things from pretty far away, and the the aforementioned object detection (all of this is just MOTION), it’s pretty amazing. In this example by fence isn’t a motion zone, it’s set to “observe” – it’s a motion sensor – that’s it, for the fence. And then porch has completely different settings as well. And again, this is ONE camera, not three, as it may appear by looking at settings alone.

Not only that, but you have at least three different Motion Detection plugins to choose from; however, ultimately, it’s not much of choice, you’re going to go with the best that your system can handle (OpenCV, PAM Diff, Webassembly, WebGL, etc), all slightly different settings and levels of power.

And then you have your Object Detection, again, unlimited zones, and they don’t have to be the same, in fact, they are completely separate, unless you want them to be the same.

You don’t HAVE to do this for every camera, but you sure can go crazy if you want to. And with scrypted NVR everyhting is 24/7, so, say goodbye to those SD cards, you don’t want them, they just get in the way.

This demo site is literally live, it’s not a software demo in the normal sense, it’s just live control over the entire app, and you can see you insanely amazing and fast the video scrubbing is, etc.

It’s a different level…