Raspberry Pi 3 B with OpenCV and Tensorflow
Our "headless" Pi will run a Python script automatically launched at boot time. This script will monitor the camera and use OpenCV to detect motion. When motion is detected, it will perform face detection on the captured frame, again using OpenCV. Detected faces will be cropped from the overall captured frame and will be saved to a USB flash drive as unlabeled data for training a neural network.
When the script detects the insertion of a second USB flash drive, it will shutdown the Pi gracefully so the full drive can be safely removed. Once powered again, it will continue to detect motion and save unlabeled data to the new drive.
Off-line, the captured faces will be manually labeled and used to design and train a neural network to identify key individuals. The trained neural network will eventually run on the Pi itself and play audible alert messages to the identified individuals.
Diagnostics information will be displayed from the Pi via a PiOLED display.
Instructions to prepare a new headless Raspberry Pi 3 from scratch.