About AI-Cube
In the AI-Cube project datacube fusion and AI-based analytics have been integrated,
demonstrated in several real-life Earth Observation application scenarios,
and deployed on a federation of CODE-DE, DIASs and further high-volume EO/geo data offerings.
Goal was to support scenarios like the following: User selects a topic
(in the project: specific crop types, specific forest types, burnt forest areas).
System determines, through a combined analysis of various large-scale data sources,
a list of regions showing the criterion selected. User gets this visualized directly
or continues analysing, possibly combining with further data sources.
AI-Cube was in part funded by the German Federal Ministry of Economic Affairs and Climate Action (BMVK). Project runtime was from September 2021 through August 2023.
Results
ML in the Datacube Engine
The rasdaman datacube engine has been enhanced so that ML models can be invoked from within datacube queries
in the OGC-standardized WCPS geo datacube query language via User-Defined Functions (UDFs) to invoke pytorch in the server for model application.
From a user (i.e., query writer) perspective this external code appears like a regular query function.
It can be used, but not downloaded so that Intellectual Property remains protected.
The following example illustrates the principle how pretrained ML models, stored in the database, can be invoked as part of a general analytics query:
for in (Sentinel_2a),
$m in (CropModel)
return encode( nn.predict( $c[...], $m ), "tiff" )
The most important advantages of this integration is that now ML is seamlessly integrated into EO datacube analytics.
For example, data can be extracted from a datacube, preprocessed as necessary, then the ML model can work on it, and the output can be processed further with all image processing and statistics functionality available.
Additionally, there is a full integration into all optimization, distributed processing, etc. the datacube engine provides already.
A particular twist of the TU Berlin contributed RSVQA technique is the integration with natural language processing:
A question is submitted along with the Sentinel patches and the model, and the output again is natural language.
The WCPS query has such a structure:
for $S1 in (S1_GRDH_IW),
$S2 in (S2_L2A)
let $patch := [ {space-time selection of 256x256 patch} ],
return
rsvqa.predict2(
$S1[subs2], $S2[subs2],
"rsvqa_trained_model.pt",
"Are there some airports?"
)
Further, the TU Berlin Begum Demir team has developed ConfigILM,
a pytorch based library that enables fast development of visual question answering systems.
Visual Question Answering (VQA) in remote sensing refers to the task of answering a question
formulated in natural language related to the semantic content of the remote sensing images.
To support development of models for this task, the Remote Sensing Image Analysis Group (RSiM)
at TU Berlin created ConfigILM, a pytorch-based library for configurable combination of pre-trained image and language models.
Publications
- P. Baumann, D. Misev: AI and Datacubes: Towards a Happy Marriage (poster). Proc. ESA Big Data from Space, Vienna, Austria, 2023
- P. Baumann, O. Campos, D. Misev: AI-Enabled Analysis-Ready Datacubes: Towards a Roadmap for More Human-Centric Services. Proc. IEEE Geoscience and Remote Sensing Society (IGARSS), July 2023, Pasadena, USA
- O.J. Campos Escobar, P. Baumann: Implementation Roadmap for Neural Networks in Array Databases. Proc. Computational Science and Computational Intelligence (CSCI), Las Vegas, USA, December 2022, doi: 10.1109/CSCI58124.2022.00024
AI-Cube Partners
Large-Scale Scientific Information Systems Group, Prof. Peter Baumann (Coordinator)
Under the lead of
Prof. Dr. Peter Baumann,
the
Large-Scale Scientific Information Systems Group
focuses on Array Databases, a research field we have pioneered since 1992:
flexible, scalable services on large, multi-dimensional arrays,
such as spatio-temporal sensor, image, simulation output, and statistics data - nowadays commonly also called "datacubes".
We address this field in all its aspects, such as algebraic modelling,
query languages, architectures, optimization, parallelization, distribution,
new hardware, application, and standardization.
Among the highlights of our research and standardization activities are
the free, open-source
rasdaman array analytics server and
the
"Big Datacube" standards in OGC, ISO, and INSPIRE.
This has been honoured by a
series of innovation awards.
Remote Sensing Image Analysis Group, Prof. Demir Begüm
The
Remote Sensing Image Analysis Group (RSIM),
led by Prof. Begüm Demir, is part of the Faculty of EECS at TU Berlin.
Our group performs research in the field of processing and analysis of remote sensing images
for Earth observation with interdisciplinary approaches associated to remote sensing,
machine learning, signal and image processing and big data management.
rasdaman GmbH
Deep-tech SME
rasdaman GmbH is a German independent academic spin-off with mission to commercializing the rasdaman ("raster data manager") datacube technology.
Its game-changing Actionable Datacubes® paradigm stands out through its flexibility, scalability, security, open standards support (such as OGC and INSPIRE),
and planetary-scale federation capabilities on Big Earth Data. The rasdaman engine is mature (TRL 9) and proven on 130+ Petabyte datacube assets,
with queries parallelized more than 1,000x in AWS; it scales seamlessly from cloud to edge, such as nanosats.
Documented through publications and patents, the rasdaman team has pioneered Datacubes and Array Databases
and remains world technology leader in high-performance analytics and fusion on spatio-temporal datacubes.
Terms of Reference
Responsible Maintainer
Constructor University Bremen gGmbH
c/o Peter Baumann
Campus Ring 12
28759 Bremen
Data Protection Regulation (GDPR)
The controller responsible for the described data collection and processing is
named in the Constructor University imprint.
When you visit our website, the Web server
temporarily stores usage information on our server
machine running in Germany) for statistical purposes
in the form of a log in order to improve the quality
of our website.
We do not transfer your personal data to third
parties without your express consent. We use
revolvermaps for determining the city of visitors
(nothing beyond this anonymized information is
available); see their privacy statement which, at
the time we evaluated it, obviously was in
accordance with GDPR.
When it comes to processing your personal data, the
GDPR grants you certain rights as a website user:
- Right of Access (Art. 15 of the
GDPR): You have the right to request
confirmation as to whether personal data
concerning you is being processed; where this is
the case, you have a right of access to this
personal data and to the information specified
in Article 15 of the GDPR.
- Right to Rectification and Right to
Erasure (Art. 16 and 17 of the GDPR): You have
the right to immediately request the
rectification of incorrect personal data
concerning you and, if necessary, the completion
of incomplete personal data. You also have the
right to request that personal data concerning
you be erased immediately if one of the reasons
listed in Art. 17 of the GDPR applies in detail,
e.g. if the data is no longer required for the
purposes for which it was collected.
- Right to Restriction of Processing
(Art. 18 of the GDPR): You have the right to
request a restriction of processing for the
duration of a review if any of the conditions
specified in Art. 18 of the GDPR have been met,
e.g. if you have lodged an objection to the
processing.
- Right to Data Portability (Art. 20
of the GDPR): In certain cases (which are
outlined in detail in Article 20 of the GDPR),
you have the right to obtain from us your
personal data in a structured, standard,
machine-readable format or to request the
transfer of such data to a third party.
- Right to Object (Art. 21 of the
GDPR): If data is collected on the basis of Art.
6(1)(f) of the GDPR (data processing on the
grounds of legitimate interests), you have the
right to object to the processing at any time
for reasons arising from your particular
situation. If you make such an objection, we
will no longer process your personal data unless
we can demonstrate compelling legitimate grounds
for the processing which override your
interests, rights, and freedoms or for the
establishment, exercise, or defense of legal
claims.
- Right to Lodge a Complaint with a
Supervisory Authority In accordance with Art. 77
of the GDPR, you have the right to lodge a
complaint with a supervisory authority if you
believe that the processing of your personal
data violates data protection regulations. This
right to lodge a complaint may particularly be
exercised before a supervisory authority in the
EU member state where you reside, work, or where
you suspect that your rights have been
infringed.