search

cvondrick / soundnet

SoundNet: Learning Sound Representations from Unlabeled Video. NIPS 2016
http://projects.csail.mit.edu/soundnet/
MIT License
462 stars 94 forks source link

All features negative #20

Open realshijy opened 4 years ago

realshijy commented 4 years ago

hi, i used trained model to extracti feature from mp3s, but all 1000 dimentional features are negative. is this normal?

MinglangQiao commented 3 years ago

I also get the same result, all features are negative