search

stanford-futuredata / noscope

Accelerating network inference over video
http://dawn.cs.stanford.edu/2017/06/22/noscope/
437 stars 122 forks source link

300+GB of memory #12

Closed GarrickLin closed 7 years ago

GarrickLin commented 7 years ago

I am confused how does this system can use up to 300+GB of memory ?

ddkang commented 7 years ago

The NoScope prototype currently loads the entire video into memory before starting downstream processing (i.e. batch processing). I'm happy to review any pull requests if you'd like to add a streaming mode.