Wyzely Detect: Adding Object Detection and Facial Recognition to (Wyze) Security Cameras

Hello, Wyze Community (and Team),

I was hoping to get your feedback on Wyzely Detect, a project attempting to use open source software to bring customizable object detection and facial recognition to security cameras, specifically Wyze Cameras.

I’m currently a high school student who started programming early in elementary school. Whilst this originally started as a facial recognition program utilizing docker-wyze-bridge to provide an RTSP stream which was then processed by ageitgey/face_recognition. I eventually added more features, such as containerization through Docker and most importantly, push-notification support through Ntfy. The most significant change I made was transitioning to Poetry for easy installation and argparse for ease-of-use. In addition, I added object recognition through YOLOv8 and switched facial recognition to serengil/deepface.

From this point on, I’ve been continuing to develop this project and my roadmap consists of adding customizable notifications on a per-camera basis in the short-term and overhauling the notification system in the long-term.

I’ve linked the demo video below. Since I’m a new forum user, I am unable to attach the file and thus, am using a YouTube video.

Feedback and questions are welcome.

Wyzely Detect Demo

2 Likes

Impressive. You have a GREAT future ahead of you! Seriously. Combining all this on your own is great. There are several others who have done similar things like Tiny Cam Pro, and Scrypted and these devs now make ongoing subscription income, some of them even with their opensource stuff because people will pay them a subscription for a few added benefits even if they could set it up all by themselves.

I have used Tiny Cam Pro and Docker Wyze Bridge, and I am getting ready to install Scrypted soon. I think it’s great that you made your own and shared it. I’m going to star it in Github and follow up on your progress over the coming time. I think this really cool.

2 Likes

This looks really cool, I will definitely check it out!

2 Likes

As I didn’t record a voiceover for the demo video, here’s a brief explanation of what was shown in the demo:

  • Detection of persons in the image of Linus Torvalds
  • Recognition of Linus Torvalds
  • Push notifications sent to Ntfy.sh with relevant content
    • Ntfy can be installed on both Android and iOS devices, in addition to having the ability to send push notifications to any compatible browser
  • Seamless switching between the detection of vehicles and the detection of people

Here’s some of the features that weren’t clearly shown in the demo, but I have implemented:

  • Support for multiple cameras
    • Any RTSP streams (including streams from docker-wyze-bridge)
    • Video capture devices (such as webcams)
  • Customization of detection thresholds
  • Customization of notification thresholds
  • Headless usage
  • Containerization of both Wyzely Detect, Ntfy, and docker-wyze-bridge to eliminate reliance on external infrastructure
    • Accomplished using Docker Compose
2 Likes

Do we have to pay for the $5 a month subscription to Ntfy, or is there a way to use it free?

Also, is there any plan to add a web UI to the system?

1 Like

Ntfy is an open source program that has a free, hosted version Ntfy.sh. Ntfy.sh has some paid features, such as being able to notify via a phone call. Thus, for most use-cases, it can be used for free, as I do in the demo video.
However, if you want greater control over Ntfy, you can self-host it (for free) easily. When using Wyzely Detect with Docker, Ntfy can also be included via Docker Compose.

Regarding the web UI, I’ve thought about it before. However, I don’t think I currently possess the skillset to make a large-scale web UI for this project, even though I would love to. However, since it’s open source, contributions are always welcome.

2 Likes