I played, more than once, a few of the sound snippets. I think the Shazam "findings" are highly inaccurate. Fun project nonetheless!
walz, could you write more about the setup, maybe to propitiate others to replicate it in other cities?
walz, could you write more about the setup, maybe to propitiate others to replicate it in other cities?
The phone records 10 minute chunks of audio at a time, in airplane mode. Every 10 minutes, airplane mode is turned off and the audio is uploaded to a server. The server then splits the audio into 15 second overlapping chunks, and each is passed to Shazam's API (no official API, but someone reverse engineered it and made a great Python package). This setup is super power efficient! The phone dips down to a minimum 70% percent battery by the early morning.
https://m.youtube.com/watch?v=YVTnj3OIhwI
I found it especially insightful because he started from the beginning and traced the thought process as the algorithm developed and became more sophisticated.
I suspect moisture damage from Karl The Fog rolling in every afternoon is more likely to kill it.
There's already people here discussing the best way to locate it. Sooner or later someone's gonna find a "free phone" and trade it for a point of meth somewhere just off 16th and Mission...
Just clicked around and you're right: the Sep 29 5:19pm snippet detected "Celebration" by Kool and the Gang, and there's almost nothing there. But it's in there.
Had I not known what I was listening for, and been intentionally listening, there's zero chance I'd have picked up on it.
It does feel like magic.
The “Not Like Us” snippet (09/29 2:43pm) is easily recognizable though. And “Rockabye” can be heard at 3:05pm.
Are you sure about that? High frequencies don't propagate as well (and, beyond a point, aren't reproduced at all by cheap speakers), so that would seem to limit its effectiveness pretty severely.