Hey all,
We've been discussing machine learning and deep learning approaches
quite often in the last couple weeks. New tools such as the ML Enabler
<https://medium.com/devseed/ml-enabler-completing-the-machine-learning-pipeline-for-mapping-3aae94fa9e94>
or the rapId editor <https://github.com/facebookincubator/RapiD> might
change the way crowdsourced data is produced in the future. I guess,
that these will also be heavily discussed during this year's HOT summit
and state of the map conference (come to Heidelberg! ;)).
Since I'm working at a research institution, I wanted to share another
piece of (scientific) work on that topic. Together with colleagues from
the Heidelberg Institute for Geoinformation Technology
<https://heigit.org/> and the GIScience Research Group
<https://www.geog.uni-heidelberg.de/gis/index_en.html> we investigated
the potential of Deep Learning in combination with MapSwipe’s
<http://mapswipe.org/> crowdsourcing approach. If this sounds
interesting to you, you can find out more at the blogpost
<http://k1z.blog.uni-heidelberg.de/2019/07/31/mapping-human-settlements-with-higher-accuracy-and-less-volunteer-efforts-by-combining-crowdsourcing-and-deep-learning/>I
wrote or directly look at our paper:
Herfort, B.; Li, H.; Fendrich, S.; Lautenbach, S.; Zipf, A. Mapping
Human Settlements with Higher Accuracy and Less Volunteer Efforts by
Combining Crowdsourcing and Deep Learning
<https://www.mdpi.com/2072-4292/11/15/1799>. /Remote Sens./ *2019*,
/11/, 1799.
Together with the broader MapSwipe community we are currently working on
new project types for MapSwipe as well. This will also allow us to
integrate machine learning results better into our existing
crowdsourcing workflows. Sounds Interesting?Check out MapSwipe’s Github
repositories <https://github.com/mapswipe> or join the MapSwipe working
group.
Have a nice day,
Benni
_______________________________________________
HOT mailing list
HOT@openstreetmap.org
https://lists.openstreetmap.org/listinfo/hot