Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla Software updates - Australia

This site may earn commission on affiliate links.
Would have thought reading and responding to speed signs is somewhat fundamental to the autonomous car that our tesla can apparently become.
I was thinking that would be an autopilot feature rather than a navigation feature. Might explain why the speed limit recognition is so far off with the maps. If they are fairly close with speed sign reading then there is not much point wasting time and money updating speed zones. Having said that Open Source Maps would seem pretty cheap.
 
After map update, and I still have a main road in Tassie (Channel Highway) which is 60km/h in both directions but K.I.T.T. says 110km/h in one direction and 40km/h in the other! I just keep reporting bugs about the speed limit and even updated www.openstreetbrowser.org as heard that might have been a source of data... Oddly enough, all my wife wants in a software update is the ability for a phone call not to scare her and allow the ringtone to be changed and she can't believe they make Backgammon harder instead. Anyway, hoping for something substantial with FSD to come through before the end of the month as toying with getting in before the FSD increase in July but if it drives 110km/h or 40km/h on a 60km/h street; then probably best to wait. (sorry - small rant in there)...
 
  • Like
Reactions: nigwil
As I understand it, they cannot use visual recognition of speed signs because they do not own the patent for it.
Perhaps, but Tweets by Elon suggest it could be something else as per: Tesla speed limit sign recognition

Third Row Tesla Podcast @thirdrowtesla May 6
Tesla is working on automatically reading speed limit signs for use with Autopilot. This should help when map data is incorrect out
This feature is actually a lot trickier than it seems, so other things had to be prioritized first. But it's coming eventually...
@elonmusk
My interpretation having watched presentations by the Tesla team working on their full-self-driving and various Elon tweets is that they want to solve the whole problem. That is using video-labelling (scene analysis so that all objects are tagged) they can direct the recognition module to the signage in the scene, rather than the simpler existing approaches where the recognition module is simple trained on "speed signage" at an image level. Tesla's video-labelling approach would be a more complete solution and handle more edge-cases.
 
  • Informative
Reactions: Hairyman
Perhaps, but Tweets by Elon suggest it could be something else as per: Tesla speed limit sign recognition


My interpretation having watched presentations by the Tesla team working on their full-self-driving and various Elon tweets is that they want to solve the whole problem. That is using video-labelling (scene analysis so that all objects are tagged) they can direct the recognition module to the signage in the scene, rather than the simpler existing approaches where the recognition module is simple trained on "speed signage" at an image level. Tesla's video-labelling approach would be a more complete solution and handle more edge-cases.
If thats the case, why are we only seeing traffic cones? Shouldn’t that be part of the same video labelling?
 
It’s similar to governments holding out with drug companies on the cost of critical drugs, in the hope of lowering the price & for how long they can handle the backlash
Tesla is holding out in the hope Mobileye will lower the royalties. Case of who blinks first
Can’t comment on the Pharmaceutical Benefits Program but with respect to Tesla And Mobileye, I doubt that There will be any collaboration going forward so I don’t think Tesla is holding out for anything.
Strangely my AP1 Model S saw a temporary roadworks speed sign and overrode the normal speed limit allowing me to use TACC to limit my speed, however in other areas where the map speed and the signposted speed disagrees the car favours the map speed, which is generally wrong.
 
  • Informative
Reactions: Hairyman