Hacker Timesnew | past | comments | ask | show | jobs | submitlogin

I don't get this. It seems to say, e.g., "We reverse-engineered the shape of Texas by drawing a line around all the photos marked 'Texas'."

So? How did the photos get marked "Texas" in the first place?



Each one of those illustrations represents the boundaries of a particular place whose outline was generated using nothing but the latitudes and longitudes of the geotagged photos associated with that location’s WOE ID.


Correct. But where does the WOE ID come from? Are people manually typing those in or they are using something like http://developer.yahoo.com/geo/ to come up with the WOE ID? If WOE ID is generated based on latitude/longitude of a place, then all they are doing is a plotting data originally retrieved from their own DB.

Say I have a table of 40,000 zipcodes in US. You take a picture in zipcode 33708 and ask me for WOE ID / state and I say "Florida" and then you upload the picture. Now do this ten million times for every zipcode. Now I will look at every picture with WOE ID / state = Florida and find a list of zipcodes and map them. Obviously they will map out the entire state of Florida. After all I gave you all the state names.

Not that this isn't cool or anything. Just trying to determine how cool it is. If they had used latitude/longitude + tags to generate this, then it would be super-cool indeed. But if they did it by the method above that I described, it is cool but it's also a big "duh."


That's because the submission title is misleading (the blog post says nothing about "reverse engineering" anything.)

Here is what flickr actually did:

"Over time this got us wondering: If we plotted all the geotagged photos associated with a particular WOE ID, would we have enough data to generate a mostly accurate contour of that place? Not a perfect representation, perhaps, but something more fine-grained than a bounding box. It turns out we can."

So yes, they're just tracing around the points in their database to see what kind of shapes they will get, for the experiment of seeing how well they could represent a given (country, state, town) by connecting coordinates of photos taken closest to the real borders of those actual regions.


If they had used latitude/longitude + tags to generate this, then it would be super-cool indeed.

This is what I thought about while reading the article, but it doesn't appear to be what they did. It would be fun to try this, though, and probably wouldn't be that hard, either.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: