openalpr/openalpr ? reverse-engineered prompt

I want to build a tool that can automatically read license plates from photos. The idea is that I can give it a picture of a car, and it will analyze it and return the characters on the plate as text. It should also provide a confidence score for its best guesses.

Let's make this a simple command-line utility. I want to be able to type a command, give it a path to an image file, and see the results printed out. It should be able to handle plates from different regions, like the US and Europe. The project also mentions it can analyze video streams, so if we can get that working too, that would be fantastic.

Please handle the full setup, including compiling the project from the source and installing any dependencies like OpenCV or Tesseract that it needs to run. The repo has a Dockerfile, so using that to build and run the tool might be the easiest approach.