Luke Ditria has created a tinier and less expensive version of a wildlife-monitoring device that he built earlier. The tool makes use of AI to detect animals in real time, without needing the internet. The original version’s use of a Raspberry Pi 5 and Hailo AI accelerator led to its increased price. This compact version uses a Raspberry Pi Zero 2 W and a smart camera that has AI built into it.
The setup makes use of a neural network model called YOLOv8N to detect animals and identify species right on the device. The use of edge AI negates the need to send data to the internet or cloud, making it a great fit for forests and other regions without proper network coverage. The infrared filter on the Sony IMX500 lets it capture true-to-life details, but limits its use to only during the daytime. A Witty Pi 4 manages battery charging and adds a real-time clock, allowing the user to decide when the device is turned on and to avoid running it pointlessly at night.
A 3D-printed case was designed for the device to keep it protected and waterproof. The current build uses basic demo code, but the creator is working on improving the software. The device can be mounted to a tree using Velcro straps. Images are stored in a USB drive every time an animal is detected, and the user can remove it at the end of the day to see the results. After a few days of testing the camera in a yard using bird seed to attract birds, magpies’ visits were captured consistently daily at around 8:00 AM.
