cea2aj/rpi-vision repository overview

⁠rpi-vision

This arm docker image runs a tensorflow object detection model from google.
It is based on this repo: https://github.com/EdjeElectronics/TensorFlow-Object-Detection-on-the-Raspberry-Pi⁠

⁠Running on a pi

The camera must be installed on the pi and enabled with raspi-config. Then run:

docker run --device=/dev/vcsm --device=/dev/vchiq -it cea2aj/rpi-vision 

⁠Output

The container takes at least 30 seconds to start up. During this time, you should see the camera on the pi turn on (the light should turn red). Once started, the output is a json string of the following format:

data = { "name", "id", "confidence", "box", "time" }

Name: Name of the object it sees. Id: Internal id of the object. Confidence: Confidence level from 0 to 1. Box: Where in the frame the object is located. Time: Timestamp of when the object was detected.