Minnesänger wrote:Basically, for those not in the know, if I jaywalked in Shenzhen the local government could, and does, automatically fine me via my ID-linked ewallet and social network accounts. No letters, no juries, just crime and fine. Within 20 seconds.
djchump wrote:There must be false positives all over the place. I’ve seen 96% and 80% false positive failure rates quoted for the London tests.
Unlikely wrote:Comprehensive isn't the word I'd use.
Minnesänger wrote:Basically, for those not in the know, if I jaywalked in Shenzhen the local government could, and does, automatically fine me via my ID-linked ewallet and social network accounts. No letters, no juries, just crime and fine. Within 20 seconds.
djchump wrote:There must be false positives all over the place. I’ve seen 96% and 80% false positive failure rates quoted for the London tests.
SpaceGazelle wrote:Getting labelled data is the tricky part because you're going to need things like social networks to get that for you, which seemingly isn't a problem in China.
GooberTheHat wrote:When you layer the data, facial recognition (from multiple sources in the area), gps data from personal devices, wifi data from connections people make as the walk around, spending data from e wallets etc you can build a pretty comprehensive pattern of life. I imagine that false positives happen much less often than they would in the UK, where thankfully our police aren't allowed to aggregate all that data
acemuzzy wrote:Can you protest auto-fines in China? Or is it just like "tough shit" if the robots get it wrong?
acemuzzy wrote:Can you protest auto-fines in China? Or is it just like "tough shit" if the robots get it wrong?
Common feature of CV CNNs/GANs is that they detect texture and are terrible at detecting silhouette/shape, due to the unsupervised learning and datasets used.SpaceGazelle wrote:... Seems it was picking up skin patterns like moles and veins and you could show it portions of faces like the top or botton and it did ok. And that's just me doing it over a couple of weeks as a noob.
It looks like you're new here. If you want to get involved, click one of these buttons!