The-Structure detection on large map images with machine learning techniques consists of three services, namely SPACE-ML CAESAR, SPACE-ML Astro ML and SPACE-ML LSE.

SPACE-ML CAESAR

CAESAR (Compact And Extended Sources Automated Recognition) service provides a straightforward solution to segment astrophysical FITS maps, allowing for the extraction and characterization of both compact (e.g. stars, galaxies) and extended sources (e.g. galactic filaments, supernovae remnants).

SPACE-ML Astro ML

These tools are Deep Learning models included and accessed by the SPACE-ML CAESAR service and also made available through the C3.1 AI Science Gateway.

SPACE-ML LSE service

Latent Space Explorer (LSE) support analysis of image datasets via unsupervised machine learning methods. It allows to extract a compact representation from data by representation learning models (e.g. autoencoders). The information extracted can be then visualized using the projector. The latter allows visualizing the data in a 2D or 3D space in an interactive fashion. The system then allows performing clustering algorithms to detect potentially relevant ways to group images and to support the definition of novel classification schemes.

The Map Making and Mosaicing of Multidimensional Space Images service delivers a user-friendly cloudbased version of the already existing workflow for map making and mosaicing of multidimensional map images based on open-source software such as Unimap and Montage.

Space science requires services for making high quality images from the raw data captured by instruments (map making) and for assembling those images into custom mosaics (mosaicking). There are some tools available for the community that are in TRL 6 and will be ported to the EOSC, optimizing the specific features as needed by the relevant scientific groups. Concerning radioastronomy data, interferometric observations, due to the peculiarities of data acquisition, the service will take into account primary beam effects as a weight for data to be mosaiced. Planetary mapping requires high image quality so that individual data products/granule are co-registered to achieve iteratively cartographic description (e.g., geologic) of a planetary surface. To achieve mapping there is need for higher-level products, that are made accessible (high-level data are not always available on public space agency archives, such as PDS/PSA) and co-registered, mosaicked as needed.

The Space-Mos provides functionalities for making high quality images from the raw data captured by instruments (map making) and for assembling those images into custom mosaics (mosaicking). These are Backend Service functionalities exposed through and accessed from the Via Lactea and ADAM-SPACE services. 

The FAIR Data Management and Visualization service provides an advanced operational solution for data management and visualization service for space FAIR data based on widespread and popular tools like VisIVO, ADN and PlanetServer. It consists of three services (SPACE-VIS Via Lactea Service, SPACE-VIS ADN Service and ADAM-SPACE).

SPACE-VIS Via Lactea Service

The Via Lactea service provides an advanced operational solution for data management and visualization of astrophysics FAIR data surveys of the Galactic Plane to study the star formation process of the Milky Way. The Via Lactea Visual Analytic (VLVA) tool combines different types of visualization to perform the analysis exploring the correlation between different data, for example 2D intensity images with 3D molecular spectral cubes. 

SPACE-VIS Astra Data Navigator Service

Astra Data Navigator (ADN) is a tool for 3D visualisation of stellar catalogues (stars, asteroids, planets, satellites, etc.) in a navigable environment. The creation of the models takes place in real time based on the physical parameters read from the catalogues to try to make the representation as realistic as possible. In addition to reading the stellar catalogues, for bodies where it is possible, the placement in the environment is carried out through the use of Spice Kernels.

ADAM-SPACE

A major obstacle in Planetary data exploration is the technical steps involved in exploring the data. The goal of this service is to handle data and provide specific products of broad use (e.g. mosaicking) as well as products that demand processing intensive tasks (e.g. landing-sites), and make those products available publicly to ease access and allow users to reuse data. ADAM Space is a thematic service for planetary data of the ADAM platform  to enable access to surface image data from Mars provided by NASA and ESA imagery.

The Air Quality Estimation, Monitoring and Forecasting service delivers a novel cloud-based solution providing crucial information and products to a variety of stakeholder in agriculture, urban/ city authorities, health, insurance agencies and relative governmental authorities.

This new EOSC service will be delivered by further optimizing, automating and validating existing software modules for air quality estimation along with the novel integration of satellite, geospatial and in-situ data. The user interface will efficiently demonstrate examples with several different challenging datasets which will target different user communities (urban air quality authorities, meteorological services, energy and power generating sector, and industrial air pollutant emitters) The examples will also demonstrate to the user the attractiveness of the package and its ease of use and also how it can be used, not only, to improve urban planning, strategy and interaction with the city residents but also can be integrated with data from other urban environment areas (for instance, mobility) and provide different functions, such as alarms management, risks and emergency situations detection, geo-referencing data, statistical analysis and life quality reports. Furthermore, a Smart Air Quality’s management platform as a SaaS platform with web responsive interfaces and REST interfaces for integration with external systems will be integrated. The dashboards on the platform will be configurable, being able to give daily, weekly and monthly reports of all the data collected by the sensing stations (even their battery status and management of the alerts regarding malfunctioning or failure). The validation will include several datasets, application cases progressing from TRL6 to TRL8.

The Atmospheric Perturbations and Components Monitoring service performs all required analytics of atmospheric and meteorological data in order to estimate possible correlations of gaseous and particulate components of the atmosphere with earthquake and volcanic processes.

The objective is to further optimize and validate the existing data from specifically set of meteorological stations in order to determine possible correlations of gaseous and particulate components of the atmosphere with earthquake and volcanic processes. For seismic processes, the correlation will be established by quantifying the variation of a specific gas (i.e., radon) and particulate matter in the lowermost atmosphere before, during and after an earthquake event. This will offer a cutting-edge service to the international fundamental requirement and discussion on the importance and reliability of such gas/particulate variations as earthquake precursors phenomena19. Regarding the volcanic processes, it is of paramount importance to quantify the input of magma-derived products in the atmosphere in the form of gases and ashes. In particular, the front-end of the service will consist of a simple, user-friendly interface that will accept the data and guide the user in obtaining the desired correlations of atmospheric components with the meteorological parameters. It will be possible to upload the data by the user from his/her own database or from an EOSC database or from a linked to EOSC database. A variety of real data/ cases will be employed for service validation progressing from a TRL6 to a TRL8 situation.

The Greenhouse Gases Flux Density Monitoring service delivers an operational workflow for estimating flux density and fluxes of gases, aerosol, energy from data obtained from specifically set meteorological stations, validated towards standardized, regularized processes.

The objective is to further optimize, automate and validate the existing software for flux density and fluxes estimations of gases, aerosol and energy from data obtained from specifically set meteorological stations, in order to integrate it as a novel cutting-edge service in EOSC. In particular, the front-end of the service will consist of a simple, user-friendly interface that will accept the data and guide the user in obtaining the desired flux densities, and energy balance results. It will be possible to upload the data by the user from his/her own database or from an EOSC database or from a linked to EOSC database. The user interface will efficiently demonstrate examples with several different challenging datasets which will target different user communities (national greenhouse gas emissions data validation, urban air quality authorities, meteorological services, and energy and power generating sector). The back-end of the service will consist of an orchestrator that will select the optimal method or combination of methods, for processing, depending on the available dataset. Two major micrometeorological pathways will be considered and integrated based on the aerodynamic gradient method and on the principle and application of the eddy covariance method. The examples will also demonstrate to the prospective users the attractiveness of the package and its ease of use. Finally, the validation towards TRL8 will include a large quantity of data and models from the community to evaluate and regulate the calculation of the above-mentioned flux densities and fluxes.

The Seabed Classification from Multispectral, Multibeam Data service delivers a user-friendly cloud-based solution integrating cutting-edge machine learning frameworks for mapping several seabed classes, validated for archeological, geo-hazards, energy, and other applications.

U3 service focuses on the efficient exploitation of recent cutting-edge multibeam echosounders which go beyond the standard ones that collect backscatter at a single predefined frequency and can acquire backscatter at multiple spectral frequencies allows the acquisition of spatially and temporarily co-registered multispectral backscatter data offering capabilities for systematic and accurate mapping of the seabed. Based on recently developed classification framework which has been already validated in relevant environment (TRL 6) and in several datasets from the international 2018 R2Sonic Multispectral Challenge, the goal is to further optimize and develop an efficient and robust multispectral, multibeam data processing system integrating advanced machine learning tools for seabed classification. In contrast to similar R&D efforts  in the proposed service different shallow and deep machine learning architectures will be further validated towards the classification of various seabed classes. An additional novelty lies in the optimization of a single classification model which will provide powerful pre-trained models for different sets of seabed nomenclatures. The validation process towards the targeted TR8 will be intensive against a number of datasets (e.g., USA, Canada, Atlantic coast, Mediterranean, Santorini, Kolumbo) demonstrating in a comprehensive way the applicability and scalability of the proposed cross-cutting service several application domains and user communities.

The Seafloor Mosaicing from Optical Data service provides an operational solution for large area representation (in the order of tens of thousands of images) of the, predominantly flat, seafloor addressing also visibility limitations from the underwater medium.

The current landscape of solutions for underwater optical mapping is still very tied to particular developments by a few research groups who create algorithms tailored to their particular acquisition hardware and target applications. As such, there is not yet any available attempt at dealing with a wide range of imaging conditions that are present in underwater surveys. The effort in service U2 is the first in its kind towards this goal powered by the cutting-edge infrastructures of EOSC. In particular, being the first of its kind, the impact of the service seems potentially significant. The key to success in this particular area is the ability to adapt to a wide range of conditions. We believe that the proposed approach, which is divided into an analysis and diagnostics task, followed by more specific processing steps, will be key to the success and adoption by a wide range of marine scientists and marine operations companies. Moreover, the target TRL is 8. The components required for this endeavour have been developed up to levels 5 (diagnostics task) and levels 6 (for the tasks of pre-processing and corrections, 2D mapping, 3D mapping). However, these have been demonstrated for particular applications and conditions, within the framework of their respective research projects. The challenge of the development behind this service is to make it robust to a wide range of input imaging conditions, which are representative of the intended science and commercial applications of the target user group. 

The Bathymetry Mapping from Acoustic Data service (UW-BAT) delivers an advanced user-friendly, cloud-based version of the popular open-source MB-System software for post-processing bathymetry through Jupyter notebooks with additional functionalities.

UW-BAT aims to create an easy to use agile and fit-for-purpose workflow of multibeam data processing as simple as upload the data, choose and apply basic filters/corrections, define the desired resolution for grids/maps and retrieve the data products from the cloud server. Furthermore, the logging of activities and the potential to share Jupyter Notebooks could aid to cooperate e.g. between industry and research. Both software required to develop this service (MB-System and Jupyter Notebooks) are already existing and available since many years and meet the TRL 6 requirement, but further development time and resources are required to piece them together to create the cloud-based service and validated with several datasets and application scenarios for the targeted TRL 8 level.