[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Package description review for CycloGraph

2012/7/23 Justin B Rye <jbr@edlug.org.uk>:
> Federico Brega wrote:
>>>   CycloGraph aims to be great at producing graphically appealing images
>>>   to document the stages of a race or to be shown on web pages and
>>>   journals.
>>> (If the journals aren't web pages, what are they?  Oh well.)
>> Do you think "newspaper" is a better term, or it's just a
>> consideration about the obsolescence of paper?
> There's a slight problem in that things can be shown "on" web pages
> and blogs and so on (and even individual newspaper pages, covers,
> etc.) but only "in" newspapers and (print) journals.  Also, the
> alternatives aren't exclusive: you might put the images on a web page
> in order to document the stages of a race.  And if I'm going to have
> to fiddle with the phrasing I want it to come out as a proper
> three-item list!  So how about boiling it down to something like
> "images to illustrate course descriptions, training blogs, and race
> reports"?

It's perfect.

> [...]
>>> I think this is trying to say:
>>>   Note that GPS tracking data is usually not enough to produce a
>>>   high-quality plot because although it contains a lot of points,
>>>   only a few really add any information to the plot.
>>> I don't understand the explanation, but then again I'm not familiar
>>> with GPS data.
>> The idea is that increasing the number of points over a certain
>> threshold will add noise instead of information.
>> My intention was to avoid the concept of noise which can be unfamiliar
>> for many people.
> Apparently it's unclear to me, too, because I would have thought the
> best way of dealing with noisy data was to use the biggest possible
> sample and take the average.

Any filter would work. Down-sapling is a very simple filter, for sure
it's not optimal with respect to the plotted profile, but in the
graphics is not limited to it.
So if you add too much points the other elements will overlap.
Even if the data were perfect showing too much tool-tips would be
confusing, so the goal is to keep the plot as simple as possible.
Down-sampling reduces both the measurement noise and the "visual"
noise due to presence of too much text.

>> I think it can be added:
>> "Using more points would introduce insignificant details, making the
>> image less clear"
> This may make sense to Christian...
> Or is this the altitude resolution problem you mention later?  I
> wouldn't have said that had anything much to do with "noise" - if it's
> that, just say:
>    Note that GPS tracking data is usually not enough to produce a
>    high-quality plot because of the resolution limits of its altitude
>    data.
> [...]
>>>> CycloGraph embedds a tool to create a kml file by drawing a
> [...]
>>>> polyline on a map (Open Street Map or Google Maps). The same tool
>>>> can also interface to Google Maps' directions.
>>> I really don't know what that's supposed to mean.  Its directions?
>>    can also interface to Google Maps' "get directions" service.
>> Do you think it is clear?
> Good, but I'd also avoid trying to possessivise "Google Maps":
>      can also interface to the Google Maps "get directions" service.
>>>> [...] On the other hand, if
>>>> they are too close the slope may be very inaccurate due to the
>>>> resolution of the altitude data.
> [...]
>>> How does that work, anyway?  Surely it's not saying that the road
>>> surface is a fractal and therefore gets longer and longer the more
>>> accurately you measure it...
>> If the distance between two points is less than the sample dimension
>> the plot looks like a staircase function.
> And a staircase function looks like a staircase!
>> In addition to that the derivative is very inaccurate ranging from 0
>> to some very high value.
> Okay, so the problem isn't that extra points make it less accurate;
> it's just that they only give an increasingly accurate picture of the
> limitations of the GPS data.  So how about:
>     On the other hand, no matter how close together they are, the slope
>     will only be as accurate as the resolution of the altitude data.

It's not the slope in general, but the gradient shown (by the way is
"gradient" the correct word?) that gets much worse.
The derivative gets as inaccurate as the ratio of the data sample
length over the requested sample length and can be pretty big.
Just to be clear in this case we're not talking about GPS data, but
altitudes measured by other satellites.

Many thanks and regards

Federico Brega

Reply to: