It reminds me how in 60s and 70s it was widely believed that the
computers will be doing text translation instead of human translators.
We realize now, fifty years later, that is actually a hard problem. And
it is still impossible to translate a novel or a poem by a computer
program alone.
I would be very surprised if digitizing landuse from satellite images
could be done by a robot. Even an automatic extraction of an image from
a background in the product photography does not work reliably. And it
is an easier task as in product photography it is possible to control
light, background color, etc. In fact it is mostly done manually even
though there are numerous programs and Photoshop plugins for it, which
kind of work in some circumstances.
New products for product photography & e-commerce will be appearing
endlessly. But we do not have that much landuse. The Earth surface will
not be growing, and there will be no other habitable planets in the near
future. In my opinion, a human, especially who knows the land, should be
participating in mapping landuse.
But certainly, if a breakthrough happens in a self-learning neural
network technology then the situation will change, and not only in
mapping and translation; it will be a new brave world.
What I would like to have however now is the better tools for landuse &
natural. For example, I would like to be able to hide in the JOSM
existing already power-lines, roads, paths, etc. in order to map
farlmland, woods, grassland, etc. Perhaps, it is possible in JOSM, but I
could not find it yet.
Best regards,
Oleksiy
On 24.12.16 18:21, Christian Quest wrote:
One example: OpenSolarMap...
We first start by crowdsourcing building roof orientations using a
very simple webapp (no need to register, open to anybody).
When enough contribution match they are considered OK (at least 3 more
than all other contributions).
Then, these contributions were used to train a neural network.
Then the nueral network was used to classify other roofs... and the
result has been put back as robot contribution to the crowdsourcing
webapp counting for 1 or 2 contributions depending on the level of
confidence (raw data is also available for download).
In all cases, there is always at least one human contribution, before
putting anything back to OSM.
It is also interesting to compare when human and robot do not agree ;)
Links...
http://opensolarmap.org/
https://github.com/opensolarmap
Next step is to use the same technique on other kind of challenges, like:
- landuse boundaries (to speedup/simplify Corine Land cover import
improvements)
- check road alignment with aerial imagery on "old" OSM traced
contributions
etc...
The potential of deep learning mixed with human contributions can give
very good things if done properly.
--
Christian Quest - OpenStreetMap France
_______________________________________________
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk
_______________________________________________
talk mailing list
talk@openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk
