These are chat archives for FreeCodeCamp/DataScience

31st
Oct 2017
test12221
@test12221
Oct 31 2017 19:45
Anyone work with Accelerometer/GPS smartphone sensor data?
evaristoc
@evaristoc
Oct 31 2017 20:03
Not me, I am afraid. Sounds like you are into something interesting. Hope to hear more from it. I am interested in working with GPS but not yet.
test12221
@test12221
Oct 31 2017 20:03
:(
Matthew Barlowe
@mcbarlowe
Oct 31 2017 20:04
What’s your problem though we still might be able to help
test12221
@test12221
Oct 31 2017 20:05
I would like to identify a persons driving style from the data..
I have collected a sample set of accelerometer data (x,y,z) and gps (lon,lat,speed,bearing)
any thoughts?
Matthew Barlowe
@mcbarlowe
Oct 31 2017 20:07
Do you have preset bins that define different styles?
test12221
@test12221
Oct 31 2017 20:08
just raw data at the moment
when I say style, I want to identify non-standard events within the data
Matthew Barlowe
@mcbarlowe
Oct 31 2017 20:10
What do you mean?
test12221
@test12221
Oct 31 2017 20:12
I assume sudden some events (brakes, sharp curves, accidents etc.) should trigger peaks in the accelerometer data..but I think I need to first orient the data to the vehicles x,y,z axis
does that make sense?
I have no preset bins yet..I think that would be a little further down the line..
Matthew Barlowe
@mcbarlowe
Oct 31 2017 20:22
Yeah if the datasets aren’t joined I would join them using I guess the car Id as the key.
Timothy Javins
@timjavins
Oct 31 2017 20:35
Uber does this for drivers. They monitor sharp acceleration and sharp braking (and likely more) and give reports to their drivers. There's also a company that uses iPhones to reconstruct incidents, using the sensor data. If I can find these, I'll share.
test12221
@test12221
Oct 31 2017 20:45
thanks man!
test12221
@test12221
Oct 31 2017 21:00
any ideas on how to orient the smartphone xyz readings to the vehicles xyz? The devices were placed randomly within the car.
Timothy Javins
@timjavins
Oct 31 2017 21:07
Use the cars acceleration and braking patterns to determine which way is up.
(Which way is forward, anyway - you know what I mean.)
I'm pretty sure the incident reconstruction company was Nexar.
http://mashable.com/2017/02/17/nexar-ai-dash-cam-uber-drivers/#Nz0BaM0hUEq1
Timothy Javins
@timjavins
Oct 31 2017 21:14
test12221
@test12221
Oct 31 2017 21:15
thanks..that sounds interesting!
The dataset download page is a bit different when you sign in with your Github account. They give you the same links, plus chop it up into parts for you.
And I'll leave it at that.
Eric Leung
@erictleung
Oct 31 2017 21:43
May be of interest to those interested in deep learning, Requests for Research, a list of problems curated by OpenAI that they think are worth working on.