The system works with AI signaling phone usage by driving.
Then a human will verify the photo.
AI is used to respect people’s privacy.
The combination of the AI detection+human review leads to a 5% false negative rate, and most probably 0% false positive.
This means that the AI missed at most 5% positives, but probably less because of the human reviewer not being 100% sure there was an offense.
Just to clarify the result: the article states that AI and human review leads to 95%.
Could also be that the human is flagging actual positives, found by the AI, as false positives.
I suspect they sent through a controlled set of cars where they tested all kinds of scenarios.
Other option would be to do a human review after installing it for a day.
The system we use in NL is called “monocam”. A few years ago it caught 95% of all offenders.
This means that AI had at most 5% false negatives.
I wonder if they have improved the system in the mean time.
We have a couple of these cameras in The Netherlands.
We found it quite intrusive to look into people’s cars. Therefore the computer will flag photos, of possible offenses, and a person verifies them.
Unfortunately the movable camera has a huge lens and it’s reported to a waze-like app before they are even finished setting it up.
The fine contains a letter, a picture and payment information. If the person really wasn’t using their phone, they can file a complaint and the fine will be dismissed. Seems pretty simple to me.
However, I have not heard any complaints about it in the news and an embarrassing amount of fines has been given for this offense.