It would be interesting to know whether the images are kept or discarded when there is no phone use detected - also what the false positive rate is, for example in Victoria (as mentioned) it is illegal to use the phone to pay in a drive through establishment, but apparently not illegal to use ones wallet - similarly is it illegal to hold ones wallet to ones ear while driving? can the camera tell the difference?
We’ve all seen people wandering over the road through phone use, but often (self included) seem to forget this is nothing new, it’s just become substantially worse. Interestingly, the degree to which the device is attached to or integral to the car apparently a determining factor in it’s perceived level of distraction … whoever came up with that pitch deserves a masters in marketing (by marketing, I mean BS).
Of course the big issue here is ‘not paying attention to the task at hand’ - inattentive driving is the potential outcome from a number of factors, including distraction, fatigue, etc - factors that have been with us as long as there have been vehicles, yet the so-called experts and our elected representatives have curiously narrow focus on extending our travel times through generic and locale non-specific speed limits and playing attention onto point source problems like this by demonising a cause and not the effect, in themselves legitimate but ignoring the real issue.
All the while the ability to get a driver licence has become much more difficult bureaucratically but arguably no more difficult from an assessable skill/education perspective - well done collective governments, you can fool most of the people most of the time …
I also agree that hands free is distracting as well. Mobile phone use in traffic should be forbidden other than if pulled up safely. The stats from research argue that to use one while driving is 400% more likely to cause a crash than if not using one, that is significant, that is 400% more than any other cause of inattention.
Fatigue causes inattention … they slice, they dice, they sell what suits their end … I rest my case
In my ideal world, learner drivers would be water-boarded or otherwise kept awake for 48 hours then put through a closed circuit extreme driving ‘experience’ so they can understand what fatigue really does to the body - it is taken seriously for truck drivers, airline pilots, Maybe a bit extreme, but we really don’t take driver training seriously (perhaps we should do the same blindfolded or with ‘other distractions’ to simulate the distraction part of inattentive) … it’s a ‘right’, just like privacy and free speech and freedom of religion, except we actually expect our right to drive to become reality.
I’m certainly think inattention is an issue, and mobile phone usage is a big contributor - but selling the problem/solution matched pair and policing it seem to have jumped outside the boundary of ‘reasonable’ - together with little if any change in the input side with training and assessment …
There is about 35% that are unknown causes that may include incidents of many of the already included reasons but may not be attributable due to lack of evidence. These stats are part of Budget Insurance’s 2019 Report at Car Accident Statistics 2019 | Car Research & Statistics — Budget Direct™
I have a story in the negative - possibly not representative of the norm - commuting one day and on a conference call for work when someone cut me off with the resulting colourful vocalisation toward the offending driver (both windows up so it was only therapeutic, it was winter in Victoria) - and the conference call went silent … I wasn’t on mute … shortly followed by my boss addressing me by name and asking if there was anything I’d like to add … he commented to me later that it was clear I wasn’t concentrating on ‘the call’ … which in hindsight is rather amusing because I had no idea what they were talking about - but I remember it was a rainy day and my commute to work was otherwise peaceful through the ranges outside the metropolis of Melbourne …
From the article itself was this “The Government said 8.5 million cars were checked during the trial period and more than 100,000 drivers were found to be using their phones illegally”
From Whichcar website “A recent sixth-month trial, by tech company Acusensus, caught 100,000 drivers red handed and using their phones illegally, from 8.5 million vehicles checked”
Computerworld “During the six-month period, the system checked 8.5 million vehicles and captured more than 100,000 drivers illegally using their handsets”
They state in their blog that illegal use of mobile devices is between 1% & 5% so the NSW trial with a slightly higher than 1% infraction rate capture is in that ballpark.
I note they say they also can log things such as speed infringement, seatbelt infringement, and unregistered vehicles (perhaps even red light infringement, or drinking while driving etc) and I wonder how long it will be before they start using those abilities to capture other offences beyond the mobile use in NSW.
So 1.2% of drivers do the wrong thing, 98.8% of drivers do the right thing - and everybody pays with ever more intrusive surveillance.
Isn’t that always the way? The law-abiding majority are punished for the sins of the idiot few.
That’s always the problem. How long before the ‘law’ requires them to retain all images for ‘2’ years ‘just in case’ someone is subsequently being investigated for a crime (that is completely unrelated to enforcing the road rules)?
I would assume that the images are manually checked i.e. computer puts up candidates for “using mobile phone while driving” - and then someone looks at it before sending out the infringement notice. We don’t want RoboMobo.
That is correct, from the FAQ that @draughtrider links to is “Every candidate image of illegal mobile phone use is presented for efficient human review against the adjudication rules of the relevant authority”. The images that are not at a satisfactory level of clarity (meaning the AI can’t determine likely use) are not passed on for human review. Also from the FAQ “comprising both the automated detection and the review service to filter out any images that don’t meet the authority’s definition of a prosecutable case[my highlighting] of illegal mobile phone use”
So based on that, the computer’s false positive rate would be interesting from a technical and administrative perspective but not from a legal perspective.
They say 95% are able to be detected so 5 of 100 positive cases are rejected as being “undetectable” to their requirements of certainty before the 95 remaining ones being sent to a human reviewer to make the final determination.
What constitutes “driving”? Looking at photo number 3 (yes, it’s grossly irresponsible but) which person is driving?
(Does the AI handle imported vehicles that are left-hand drive?)
One could suggest that the whole thing is a big waste of money because in not too many years all of the photographed cars (urban, motorway) will be self-driving and the law will change so as not to require any person to be involved in the driving (or even be present in the car). They should put the money towards that, which will solve “idiots using mobile phones while driving” and “idiots who are fatigued / drunk / drugged” and “idiots who are just careless / inattentive” - as well as dealing Australia in to future industries.
The ‘authorities definition’ could easily be anything they determine, including ‘everything is an infraction unless they complain’ … I suspect the ‘authority’ would also have input into the definition of how clear the image needs to be. One thing that is clear - the images need to form part of the communication with the customer …
from the FAQ …
Sophisticated algorithms automatically detect the driver and determine the likelihood of phone use. This detection can be performed in real time to assist Police operations.
Integral to the program is the privacy of data. In the automated solution licence plate data remains always encrypted and only viewable by the authority, and human reviewers are presented with images only of the drivers, not passengers or licence plates. Acusensus is also able to automatically blur out everything in extracted images bar the driver if required or requested.
It’s not hard to see where this could go. I wonder if the software is smart enough to pick up left hand drive or dual control vehicles? Words like “always encrypted” and “able to” don’t inspire any confidence.
The solution has been trained and tested across millions of vehicle transits and manually verified to successfully detect at the highest accuracy levels.
I wonder where all that data is currently
The system operates successfully on vehicles travelling up to 300km/h without motion blur and without distortion.
Well tested at those speeds? Using a mobile was the last thing on my mind when I drove at those kinds of speeds (legally).
That also doesn’t inspire any confidence. “required” or “requested” by whom? The government when needed to avoid embarrassment for themselves? The government but they will never make such a request?
The Minister actually said about self driving vehicles but in the meantime he is trying to reduce deaths, accidents and injuries related to the use of mobile devices. That’s what he says and I have no proof of otherwise.
I don’t know what the AI is able to discern but my reading so far indicates it looks for information in a car to determine whether it is a driver or a passenger, this may include whether there is a steering wheel, but you can contact the company and they may be able to provide some more information about that.
Basically if you don’t use a phone you won’t be booked, photoed doing so, or required to pay the fine. Pretty simple way to avoid the issue of being penalised for breaking the law.
In reply to my query the Managing Director of Acusensus Pty Ltd, Alexander Jannink, responded about detecting RHD & LHD usage that the system does detect both.
