Re: [HOT] Terrapattern: query-by-example for satellite imagery

[Christian Ledermann]

also of interest in this context:
https://github.com/trailbehind/DeepOSM

DeepOSM

Detect roads and features in satellite imagery, by training neural
networks with OpenStreetMap (OSM) data. This code lets you:

Download a chunk of satellite imagery
Download OSM data that shows roads/features for that area
Generate training and evaluation data


On 27 May 2016 at 01:48, Alan McConchie  wrote:
> There's at least one problem with using Terrapattern as-is: it's using Google 
> imagery, so therefore anything you derive from Terrapattern wouldn't be 
> license-compatible with OSM.
>
> Now, if they hooked it up to Mapbox Satellite...
> ___
> HOT mailing list
> HOT@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/hot



-- 
Best Regards,

Christian Ledermann

Newark-on-Trent - UK
Mobile : +44 7474997517

https://uk.linkedin.com/in/christianledermann
https://github.com/cleder/


<*)))>{

If you save the living environment, the biodiversity that we have left,
you will also automatically save the physical environment, too. But If
you only save the physical environment, you will ultimately lose both.

1) Don’t drive species to extinction

2) Don’t destroy a habitat that species rely on.

3) Don’t change the climate in ways that will result in the above.

}<(((*>

___
HOT mailing list
HOT@openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot

Re: [HOT] Terrapattern: query-by-example for satellite imagery

[Alan McConchie]

There's at least one problem with using Terrapattern as-is: it's using Google 
imagery, so therefore anything you derive from Terrapattern wouldn't be 
license-compatible with OSM.

Now, if they hooked it up to Mapbox Satellite...
___
HOT mailing list
HOT@openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot

Re: [HOT] Terrapattern: query-by-example for satellite imagery

[Hakuch]

On 26.05.2016 17:40, Blake Girardot HOT/OSM wrote:
> This discussion leads to a lot of potential ideas.
>
> they mention OSM in text, but it was not clear to me how OSM data was
involved.

"We trained a 34-layer DCNN using hundreds of thousands of satellite
images labeled in OpenStreetMap, teaching the neural network to predict
the category of a place from a satellite photo."

the osm data was just to connect the image to actual data, so their
software could learn how an airport normally looks like

(sorry, I sent email not to list by mistake)


0x3CBE432B.asc
Description: application/pgp-keys
___
HOT mailing list
HOT@openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot

Re: [HOT] Terrapattern: query-by-example for satellite imagery

[Blake Girardot HOT/OSM]

The about page is long, but well worth a full read to the end:

http://www.terrapattern.com/about

OSM data played a real role in the project, and that means everyone
here played a real role.

It is intended as a humanitarian project and seems to touch on a lot
of the issues humanitarian response and mapping faces on a regular
basis. It is pretty nice it is all open source. I am sure OSM and HOT
will be reaching out to them to see how we can all work better
together.

The Knight Foundation sponsored it in part, they support the most
amazing projects.

Please feel free to keep sharing ideas after reading their about page.

Thank you martind for putting the link in IRC and email :)

Regards
Blake


Blake Girardot
Humanitarian OpenStreetMap Team
Vice President, HOT Board of Directors
skype: jblakegirardot
HOT Core Team Contact: i...@hotosm.org


On Thu, May 26, 2016 at 5:09 PM, joost schouppe
 wrote:
> It is not, at least not at this time. For example, I tried finding airports
> by clicking on a plane. It returns some planes and a lot of asphalt. It
> might however be useful to quickly identify a very rough outline of built up
> land or roads. In my case: airports did show up as hotspots. So it might be
> useful to direct mapping activity.
>
> 2016-05-26 16:51 GMT+02:00 Paul Uithol :
>>
>> That sounds quite interesting - I'm wondering if it could be accurate
>> enough to use as a tool for automatically tracing/classifying/tagging
>> features based on imagery in some scenarios? I imagine the error rate would
>> be quite a lot higher than with proper mapping, but in some situations
>> having a fast initial assessment (before focusing on accuracy) is useful?
>>
>>
>> On 26-05-2016 12:57, Martin Dittus wrote:
>>
>> Made the rounds on social media yesterday, I imagine a few of you have
>> already seen it.
>> http://www.terrapattern.com/
>>
>> It’s a visual search for aerial imagery — "a prototype for helping people
>> quickly scan extremely large geographical areas for specific visual
>> features”. You mark a desired feature, it finds all other occurrences.
>>
>> Their case studies mention OSM as a source data set, but they don’t yet
>> appear to be talking to anyone at HOT? "Terrapattern was only possible due
>> to the astonishing crowdsourced mapping effort of the OpenStreetMap project,
>> which has generously categorized large parts of the world with its Nominatim
>> taxonomy”.
>>
>> m.
>> ___
>> HOT mailing list
>> HOT@openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/hot
>>
>>
>> ___
>> HOT mailing list
>> HOT@openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/hot
>>
>
>
>
> --
> Joost @
> Openstreetmap | Twitter | LinkedIn | Meetup
>
> ___
> HOT mailing list
> HOT@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/hot
>

___
HOT mailing list
HOT@openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot

Re: [HOT] Terrapattern: query-by-example for satellite imagery

[Blake Girardot HOT/OSM]

This discussion leads to a lot of potential ideas.

they mention OSM in text, but it was not clear to me how OSM data was involved.

About the only feature I can think of is airports, but I could almost
see it using the current tagged airports or helipads (real heli pads,
with an H) as examples to then search a much much larger area for
similar features to review.

At the very least it would be sweet if it could use the osm data to
remote control to josm to look at those example areas and all the
matching areas it found.

or just highlight some tight rectangles or circle around the aerial
imagery of examples of a thing to search for  (heli pads for example)
and let it find everything like that, click, josm to review

I am sure I am not explaining that very well.

Probably not possible to search for radio or transmission towers or
masts, but it would very impressive if it did with a high accuracy.

Regards,
Blake


Blake Girardot
Humanitarian OpenStreetMap Team
Vice President, HOT Board of Directors
skype: jblakegirardot
HOT Core Team Contact: i...@hotosm.org


On Thu, May 26, 2016 at 5:09 PM, joost schouppe
 wrote:
> It is not, at least not at this time. For example, I tried finding airports
> by clicking on a plane. It returns some planes and a lot of asphalt. It
> might however be useful to quickly identify a very rough outline of built up
> land or roads. In my case: airports did show up as hotspots. So it might be
> useful to direct mapping activity.
>
> 2016-05-26 16:51 GMT+02:00 Paul Uithol :
>>
>> That sounds quite interesting - I'm wondering if it could be accurate
>> enough to use as a tool for automatically tracing/classifying/tagging
>> features based on imagery in some scenarios? I imagine the error rate would
>> be quite a lot higher than with proper mapping, but in some situations
>> having a fast initial assessment (before focusing on accuracy) is useful?
>>
>>
>> On 26-05-2016 12:57, Martin Dittus wrote:
>>
>> Made the rounds on social media yesterday, I imagine a few of you have
>> already seen it.
>> http://www.terrapattern.com/
>>
>> It’s a visual search for aerial imagery — "a prototype for helping people
>> quickly scan extremely large geographical areas for specific visual
>> features”. You mark a desired feature, it finds all other occurrences.
>>
>> Their case studies mention OSM as a source data set, but they don’t yet
>> appear to be talking to anyone at HOT? "Terrapattern was only possible due
>> to the astonishing crowdsourced mapping effort of the OpenStreetMap project,
>> which has generously categorized large parts of the world with its Nominatim
>> taxonomy”.
>>
>> m.
>> ___
>> HOT mailing list
>> HOT@openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/hot
>>
>>
>> ___
>> HOT mailing list
>> HOT@openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/hot
>>
>
>
>
> --
> Joost @
> Openstreetmap | Twitter | LinkedIn | Meetup
>
> ___
> HOT mailing list
> HOT@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/hot
>

___
HOT mailing list
HOT@openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot

Re: [HOT] Terrapattern: query-by-example for satellite imagery

[joost schouppe]

It is not, at least not at this time. For example, I tried finding airports
by clicking on a plane. It returns some planes and a lot of asphalt. It
might however be useful to quickly identify a very rough outline of built
up land or roads. In my case: airports did show up as hotspots. So it might
be useful to direct mapping activity.

2016-05-26 16:51 GMT+02:00 Paul Uithol :

> That sounds quite interesting - I'm wondering if it could be accurate
> enough to use as a tool for automatically tracing/classifying/tagging
> features based on imagery in some scenarios? I imagine the error rate would
> be quite a lot higher than with proper mapping, but in some situations
> having a fast initial assessment (before focusing on accuracy) is useful?
>
>
> On 26-05-2016 12:57, Martin Dittus wrote:
>
> Made the rounds on social media yesterday, I imagine a few of you have 
> already seen it.http://www.terrapattern.com/
>
> It’s a visual search for aerial imagery — "a prototype for helping people 
> quickly scan extremely large geographical areas for specific visual 
> features”. You mark a desired feature, it finds all other occurrences.
>
> Their case studies mention OSM as a source data set, but they don’t yet 
> appear to be talking to anyone at HOT? "Terrapattern was only possible due to 
> the astonishing crowdsourced mapping effort of the OpenStreetMap project, 
> which has generously categorized large parts of the world with its Nominatim 
> taxonomy”.
>
> m.
> ___
> HOT mailing 
> listHOT@openstreetmap.orghttps://lists.openstreetmap.org/listinfo/hot
>
>
> ___
> HOT mailing list
> HOT@openstreetmap.org
> https://lists.openstreetmap.org/listinfo/hot
>
>


-- 
Joost @
Openstreetmap  |
Twitter  | LinkedIn
 | Meetup

___
HOT mailing list
HOT@openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot

Re: [HOT] Terrapattern: query-by-example for satellite imagery

[Paul Uithol]

That sounds quite interesting - I'm wondering if it could be accurate 
enough to use as a tool for automatically tracing/classifying/tagging 
features based on imagery in some scenarios? I imagine the error rate 
would be quite a lot higher than with proper mapping, but in some 
situations having a fast initial assessment (before focusing on 
accuracy) is useful?


On 26-05-2016 12:57, Martin Dittus wrote:

Made the rounds on social media yesterday, I imagine a few of you have already 
seen it.
http://www.terrapattern.com/

It’s a visual search for aerial imagery — "a prototype for helping people 
quickly scan extremely large geographical areas for specific visual features”. You 
mark a desired feature, it finds all other occurrences.

Their case studies mention OSM as a source data set, but they don’t yet appear to be 
talking to anyone at HOT? "Terrapattern was only possible due to the 
astonishing crowdsourced mapping effort of the OpenStreetMap project, which has 
generously categorized large parts of the world with its Nominatim taxonomy”.

m.
___
HOT mailing list
HOT@openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot
___
HOT mailing list
HOT@openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot

[HOT] Terrapattern: query-by-example for satellite imagery

[Martin Dittus]

Made the rounds on social media yesterday, I imagine a few of you have already 
seen it.
http://www.terrapattern.com/

It’s a visual search for aerial imagery — "a prototype for helping people 
quickly scan extremely large geographical areas for specific visual features”. 
You mark a desired feature, it finds all other occurrences.

Their case studies mention OSM as a source data set, but they don’t yet appear 
to be talking to anyone at HOT? "Terrapattern was only possible due to the 
astonishing crowdsourced mapping effort of the OpenStreetMap project, which has 
generously categorized large parts of the world with its Nominatim taxonomy”.

m.
___
HOT mailing list
HOT@openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot