News for hackers. Stuff that matters.
Re:End to End Learning for Self-Driving Cars [pdf] by andrewchambers on Monday May 09, 2016 @07:03PM (#11663948)
They need some sort of prefilter to remove any bad driving habits before training begins I suppose.
Re:End to End Learning for Self-Driving Cars [pdf] by jonny_eh on Monday May 09, 2016 @07:04PM (#11663957)
I imagine with enough training data from enough drivers that bad habits will disappear as noise.
Re:End to End Learning for Self-Driving Cars [pdf] by msandford on Monday May 09, 2016 @07:11PM (#11663996)
Depends on the "bad habit". A huge number of people speed. How's that going to disappear as noise?
Re:End to End Learning for Self-Driving Cars [pdf] by andrewchambers on Tuesday May 10, 2016 @02:09AM (#11665431)
wouldn't it just converge on an average driver? not even a good one.
Re:End to End Learning for Self-Driving Cars [pdf] by cwe on Monday May 09, 2016 @07:11PM (#11663997)
I'd like to believe that would be the case, but I live in Seattle.
Re:End to End Learning for Self-Driving Cars [pdf] by johntb86 on Monday May 09, 2016 @08:56PM (#11664515)
From the paper:

To train a CNN to do lane following we only select data where the driver was staying in a lane and discard the rest.