J. Pery

Hi there! If you want we can use the workaround that @zaccharieramzi found, as it improves up to 10 times the actual implementation, at least for my use case, until...

Hi! First of all, thanks for your PR! I think you are trying to avoid the error "ERROR on binding:Address already in use.". Is that it? As they say at...

Hi, I haven't tested it yet on Windows and I don't know if it'll work due to I'm using some *nix specific libraries (like sys/socket.h or pthread.h). You could be...

As you can see in the attached graphic at https://github.com/worldveil/dejavu/pull/148#issuecomment-403382614 better than 1.2 should be enough to validate the identification. It depends on your use case, the lower the value...

You should use numpy==1.13.3 to avoid this problem

Do you mean 200 mil fingerprints or 200 mil songs? 200 milion songs will mean about 600 TB if my calcs are not wrong (~3MB per song * 200M songs)....

In #147 you can find a temporal solution

As I've checked, the most time of the recognition is spent in the fingerprint of the song to recognize, so in many cases it doesn't matter the size of your...

The main thing here is the "scipy.ndimage.filters -> maximum_filter" takes so long and there is no way to improve it without touching scipy code. You can find the function [here](https://github.com/scipy/scipy/blob/8dba340293fe20e62e173bdf2c10ae208286692f/scipy/ndimage/src/ni_filters.c#L529).