Uber’s self-driving car could not detect pedestrians outside of a crosswalk

Re:Uber’s self-driving car could not detect pedestrians outs... by jacquesm on Wednesday November 06, 2019 @06:32PM (#21468754)

> Classification is not needed to avoid objects.

Taking your disclaimer into account, that is a very wrong assumption. If you classify an object wrongly that means your model will not be able to properly estimate what the object's capabilities are, including its potential acceleration speed and most likely trajectory from a stand-still. And that information will come in very handy in 'object avoidance' because a stationary object could start to move any time.

[ Parent | on HN ]

Re:Uber’s self-driving car could not detect pedestrians outs... by a_t48 on Wednesday November 06, 2019 @07:54PM (#21469370)

Yes, I suppose I wasn't exact when I stated that. Classification helps very much, but is no excuse to throw away other data you have. It should be an augmentation to the track the object has.

[ Parent | on HN ]

Re:Uber’s self-driving car could not detect pedestrians outs... by jacquesm on Wednesday November 06, 2019 @08:04PM (#21469427)

Then we are in agreement.

[ Parent | on HN ]