neat-EO.pink

Efficient #AI4EO OpenSource framework, powered by @data_pink

[neat-EO.pink @neat_eo]

@GFDRR Up to date URL:

https://t.co/E0gryH3q7p

[neat-EO.pink @neat_eo]

https://t.co/fFflOgyMIE 101 tutorial was updated,
to become even more simple:

https://t.co/bMrg71LTS1

You only need ~2h, a GPU server and copy/paste skills...

Thanks to @GFDRR for the new training dataset

[neat-EO.pink @neat_eo]

Worth to mention, that https://t.co/fFflOgyMIE project use now his own self hosted gitea, rather than former Github + PyPi.

US EAR related.

Authentication is based on GitHub credentials, for an easy sign-in.

neat-EO 0.7.4 just released.

Thanks @giteaio team !

[neat-EO.pink @neat_eo]

https://t.co/tdc9592C2R 0.6.0 release with:
- MultiClasses support
- MultiGPUs @PyTorch DDP train
- Metrics improvments
- Latest @PyTorch 1.4
- Refactor and bugfixes

#ComputerVision framework for #GeoSpatial Imagery

#AI4EO #FOSS4G

[neat-EO.pink @neat_eo]

@mattdance https://t.co/tdc9592C2R allows you, to do it by yourself.

It clearly been designed for such usages.
You could give a look on the online tutorials...

And if you prefer to contract someone @data_pink can handle it.

[neat-EO.pink @neat_eo]

Next coming https://t.co/tdc9592C2R presentation @OSS_Paris, December 11, in the #AI main track.

OpenSource #ComputerVision framework for #GeoSpatial Imagery, at scale.

https://t.co/YK9SrE7NXR

See you there !

[neat-EO.pink @neat_eo]

Zang Bin gathered and maintains an impressive
#EO #OpenDataSet list:

https://t.co/16XhWPRFn2

cc @chrieke

[neat-EO.pink @neat_eo]

@zpnca @o_courtin @RoboSatPink 101 Workshop given @foss4g :

https://t.co/q6u0R6GL1Y

Pictures are dynamic, click on them :
- Pink: predicted by the model
- Green: in labels or OSM
- Grey: both agree
- Pink squares: significant differences between both

Whole stuff took ~1h30 only on a single GPU

[neat-EO.pink @neat_eo]

@zpnca @CosmiqWorks Yes ability to automagically and statically render each step of the process is that helpful.

About projection, not sure if it was clear enough,
but you can use any projection for inputs (raster or vector).

Just, they will be transparently reprojected in EPSG:3857.

[neat-EO.pink @neat_eo]

"This is the beginning of a beautiful friendship"

@RoboSatPink 101 Tutorial, given @foss4g:

https://t.co/q6u0R6GL1Y

#PinkMagic #EO #GIS #DeepLearning

[Anze Zupanc @zpnca]

@RoboSatPink @o_courtin Mission accomplished. Successfully trained a building detector with @RoboSatPink. Next step: use #eolearn to prepare/filter Sentinel-2 imagery and labels and use @RoboSatPink to develop forest detector for example. Thanks @o_courtin for excellent tutorial at #foss4g2019 👏👏👏

[neat-EO.pink @neat_eo]

Coming https://t.co/tdc9592C2R rendez-vous:

- 08/26 Workshop @foss4g
https://t.co/0nDr9pOvC4

- 08/29 Talk @foss4g
https://t.co/6lyh9eoCmG

- 09/21 Talk @sotm
https://t.co/Y7OBhNGv7v

#OpenSource #ComputerVision #Ecosystem for #GIS #Imagery

[neat-EO.pink @neat_eo]

Coming https://t.co/O2w3Lf3Mmv presentations:

- May 29 #NuitCodeCitoyen @Tuba_Lyon
https://t.co/2ET1v8IZVf

- June 11 & 12 @AI_PARIS_ on @data_pink booth
https://t.co/sX3akfhWdX

- June 13 @ow2
https://t.co/aXYo6xLtVR

#ComputerVision ecosystem for #GeoSpatial #Imagery

[neat-EO.pink @neat_eo]

@mvexel @bhousel @iandees @waldojaquith @mvexel
Don't hesitate to report any documentation lack, from a new user point of view.

Or to use Gitter if you need some assistance: https://t.co/r4RXB9agSV

And FYI there's already a sample Osmium Parking Handler example,
to generate labels from OSM: https://t.co/Ink8l2NXqX

[neat-EO.pink @neat_eo]

The improved results with latest https://t.co/tdc9592C2R release:
https://t.co/R2mFPpG18R

Pink: predicted by model
Green: in putative labels
Grey: both match
Square: significant differences between both

From #OpenData to #OpenDataSet tuto:
https://t.co/SSx3friqVo

[neat-EO.pink @neat_eo]

From #OpenData to #OpenDataSet, with https://t.co/tdc9592C2R

Pitch: How to qualify a decent #ComputerVision #AerialImagery training DataSet, from @grandlyon #OpenData.

https://t.co/SSx3friqVo

[neat-EO.pink @neat_eo]

https://t.co/tdc9592C2R 0.4.0:

ChangeLog:
- Albumentation #DataAugmentation
- Improve performances
- @PyTorch JIT export
- Enhance CLI, doc and WebUI

Install: pip3 https://t.co/tdc9592C2R
Chat: https://t.co/r4RXB9agSV

#SemanticSegmentation ecosystem for #GeoSpatial #Imagery

[neat-EO.pink @neat_eo]

https://t.co/tdc9592C2R 0.3.2 release:

- @postgis support for rasterize tool
- MultiBands support for predict tool
- PyPI install: pip install https://t.co/tdc9592C2R
- Latest @pytorch support
- Few BugFixes and refactor

#SemanticSegmentation ecosystem for #GeoSpatial #Imagery

[neat-EO.pink @neat_eo]

How to improve @openstreetmap #DataQuality with #DeepLearning ?

Spoiler: with https://t.co/O2w3Lf3Mmv

@fosdem slides: https://t.co/jZTwaCiIAJ

[neat-EO.pink @neat_eo]

https://t.co/tdc9592C2R 0.2.0 release, with:

- Features extraction
- Models and losses extension
- Pytorch 1.0 support
- Bug Fixes and cleanup

[DataPink @data_pink]

In coming days, two events to see latest https://t.co/3Dv3Be8fIg progress:

24-25 Jan
@dataspacexyz on @data_pink booth
https://t.co/O9cuE4A5pU

03 Feb
@fosdem on Geospatial devroom
https://t.co/d4Mh47Mp6i

@RobosatP #SemanticSegmentation ecosystem for #GeoSpatial #Imagery

[neat-EO.pink @neat_eo]

@mouthofmorrison @bhousel @grandlyon And any suggestion to coordinates efforts on #OpenDataSet among #GIS Company already involved in #ComputerVision, as @azavea, @developmentseed or @mapbox is really welcome...

[neat-EO.pink @neat_eo]

@mouthofmorrison @bhousel @grandlyon I will be, what i will be.

Indeed supervised learning progress is driven by #OpenDataSet availability.

And models, losses and metrics, have to be switchable to allow anyone to use only those he really need, and to bring his owns.

Pre-trained models naturally follows then...