YoloCam

Introduction.

YoloCam is a software package running on a Raspberry Pi 4, 3 or Zero 2 W. It transforms the Rpi into a stand-alone AI-powered camera. No expensive cloud services are required. With a deep learning model, it detects objects in the camera scene. You can define what actions YoloCam performs when it recognizes an object. For instance, send you an email. Or make a movie and store it at Gdrive. Or activate one of its GPIO pins. At the same time, you can view your footage in any browser. Now, it is your chance to give your obsolete RPi 3B a second life!

How it works.

The heart of the app is a state-of-the-art deep learning model. It can recognize 80 different objects in a scene. Once an object is found, it is investigated further. What is its size? Does it move? How fast?

All this information is used to trigger events. You define how the app will respond to these events in a settings file. What must it do when the postman rings at your door? Or when your dog is chewing on your slipper. Or when the neighbour's cat sneaks towards the birdcage. No more false alarms from the shadows of swaying trees or moving traffic in the background.

Flavours.

YoloCam comes in three flavours. While all have the same deep learning heart, the in or output differ.

• GPIO. A trigger sets/resets a specified GPIO pin or sends a signal to a URL.
• Email. A trigger generates emails and store event movies.
• YoloIP. It connects to the IP of commercial surveillance cameras.

GPIO.

This version activates the GPIO output pins in case of an event. There is a live feed to your browser. The GPIO outputs act in real time; there is no latency. Finally, the GPIO version can trigger URLs.

Email.

This version sends emails and records movies. There is a live feed to your browser. It has a latency of 10 seconds due to the HLS streaming. Thanks to this latency, you receive your emails 5 seconds before the actual movement is visible in your browser so you can log in. Also, the saved movies have a preview of a few seconds before the actual event takes place. Movies might stored on a USB stick or Google Drive.

YoloIP.

YoloIP connects to RTSP streams of IP cameras. Most commercial surveillance cameras use this protocol. You enter the IP address, and YoloIP will monitor the footage. More than one camera can be monitored simultaneously without latency.

YoloIP activates GPIO pins. At the same time, it generates emails. A thumbnail overview is available in your browser.

Get started.

The first step is to download an SD image of the YoloCam you prefer. Go to one of the GitHub repos given in the links at the end of this article. Here, you find the download matrix for your Pi.

The next step is flashing the image on an SD card with the balenaEtcher or the Raspberry Pi Imager. Insert your fresh-baked SD card in the Raspberry PI, and power up.

YoloCam works with a one-time license linked to the SD card. For the price of a coffee, you have your license code. Enter the code in the Key file, and you're done!

Wiki.

The wiki pages on GitHub cover all the key topics. Such as how to set the triggers, how to send emails, how to save videos to Google Drive, etc. YoloCam has many settings. It would take us way too far to cover all of this here. Please read them.

Cooling.

You must cool the Raspberry Pi. Running a deep learning model requires a lot of computer power, around 80% CPU usage. Your Pi will get very hot without proper cooling.

Software.

The application is written in C++. All code is available on the SD. You can alter the app to your needs. Two remarks. You can not modify the deep learning model. You cannot employ a user-trained trained model. Secondly, your license is linked to the SD card. If you copy the app to another card, you require a new license key for this card.