News for hackers. Stuff that matters.
Re:Show HN: A neural net that critiques your logo by eps on Thursday August 31, 2017 @02:07PM (#15142398)
This was on reddit a couple of days ago and the thread was full of examples of bland logos that scored really well and good or notable logos that scored very poorly.

I wouldn't take this neural net output too seriously.

The training size is 1 mil logos - that alone makes one wonder how this mil was ranked in the first place to separate good logos from the junk that is bound to comprise a pool of logos that huge.

Re:Show HN: A neural net that critiques your logo by Jack000 on Friday September 01, 2017 @02:48PM (#15151392)
if the tool told you something you already knew - that notable logos are good, what would be the point?

maybe there's more value in discussing why the algorithm thinks bland logos score higher than the notable ones.