BigDataGrapes aims to support all European companies active in two key industries powered by grapevines: the wine industry and the natural cosmetics one but also provide innovative tools to the food safety market. It will help them respond to the significant opportunity that big data is creating in their relevant markets, by pursuing two ambitious goals:

  • To develop and demonstrate powerful data processing technologies that will increase the efficiency of companies that need to take important business decisions dependent on access to vast and complex amounts of data. To this end, one key outcome aims to be a set of data processing tools and methods that are produced by EU technology partners, rigorously tested on data challenges informed by the grapevine-powered industries, but also transferable into other industries within the agriculture, food and beauty sectors.
  • To catalyse the creation of a data ecosystem and economy that will increase the competitive advantage of companies that serve with IT solutions to these sectors. To this end, another key outcome aims to be a data marketplace for grapevine-related data assets that will help companies and organisations evolve methods, standards and processes to help them achieve free, interoperable and secure flow of their data.

By providing tools to the IT and food companies we also created higher impact to the societies by addressing societal challenges like the food security, the resource efficiency, the water availability, the grapevine by-products biological efficacy and the food pricing.

During the life of the project a series of innovations were developed. A short description for each innovation is presented in the folowing tables:


Innovation 1

Title: Visual Decision Support System for Machine Learning model selection

Description: AHMoSE (Augmented by Human Model SElection), is a visual Decision Support System (DSS) allowing viticulture experts with little to no ML (machine learning) experience to answer critical questions in viticulture (e.g. “Which ML model should I use with this grape variety specific data”). AHMoSE - based on model-agnostic ML interpretation methods - compares and explains the predicted outcomes of various ML models and helps domain experts to select the models that fit their knowledge.


Innovation 2

Title: Correlation visualization for maximizing the predictive performance of Machine Learning models

Description: GaCoVi (Gapped Correlation Visualisation) is an innovative system for correlation visualization. It is designed to put viticulture experts in the loop of the feature selection process, as a preliminary stage to the Decision Support System (AHMoSE). GaCoVi maximizes the predictive performance of Machine Learning models by supporting: 1) the identification of irrelevant features and 2) the identification of highly correlated feature pairs and selection of the most relevant feature within them.


Innovation 3

Title: Customizable alarms for thermal and water stress thresholds

Description: A customizable alarm functionality, integrated in the Abaco Farmer application, for automatic alert on the achievement of thermal and water stress values on a specific crop parcel. For each type of alarm, the user can define: the monitoring data sources (raster data or specific sensor data); two level thresholds (user defined danger and alert alarms); the period in which the alarms must be reported and the parcels that must be monitored.


Innovation 4

Title: Adjustable colour scale for Satellite Indexes images

Description:  In order to be representative of the potential issues in the field, the Satellite Indexes Images colour scale must be adapted to the crop varieties. Geocledian and Abaco set up an API that allows to select the proper colour scale in function of the crop in the parcel. Related to the 2 piloting farms of the project the grapevine colour scale for NDVI index has been implemented. 


Innovation 5

Title: A scalable and open big data collection and processing workflow

Description:  Development of an innovative data collection and processing workflow for the food industry that utilizes big data technologies, artificial intelligence, text mining and automatic text translation methods to harmonize highly heterogeneous and dynamic data. Development of semantic vocabularies for specific domains to enable the interconnection of different data types. Development of a Data API, allowing to execute smart queries over a large number of different data types and millions of records.


Innovation 6

Title: Automated assessment and prediction of food risks

Description: Development of an automated risk assessment approach for food safety and fraud issues, that reduces the manual work needed by the food experts to identify and assess risks in ingredients used in the grapevine industry. Development of innovative predictive analytics dashboard that predicts global risks for their critical ingredients & finished products. The prediction models are specifically trained and used for each food company and are linked with the big data platform.


Innovation 7

Title: Methodology for solving business challenges in food safety using Artificial Intelligence

Description:  We developed a methodology on how to solve real business challenges the food industry is facing, through Artificial Intelligence (“Food Safety intelligence equation”). It includes steps about: a) understanding critical-business decisions, b) selecting & preparing the right data, c) selecting the right algorithm for accurate predictive analytics, d) selecting the metric for effective analytics and e) assessing the operational requirements. Α series of videos is available at the project’s website.


Innovation 8

Title: Data marketplace for grapevine-related data assets

Description:  Development of an innovative data marketplace where grapevine-powered and food industry data assets can be shared and exchanged by companies and organisations responsible for them. A number of data standards adopted, implemented and revised to facilitate free, interoperable and secure data flows. The data marketplace that is based on innovative business models that can enable the exchange and trade of data.


Innovation 9

Title: Vex: a data visualisation platform to explore heterogenous datasets

Description:  Exploring large, unorganised and heterogeneous datasets can be a challenging task. To overcome this, we have designed Vex, a simple-yet-powerful reactive web application that incorporates a collection of reusable and simple-to-use data visualisation components which can together accommodate various forms of data. Vex allows interaction inside the visualisation components to filter and sort information and customization with colours, filtering and basic operations (sum, median, mean, min, max).


Innovation 10

Title: New Indexing/Compression Techniques for Big Data

Description:  In BigDataGrapes, we developed new data representations to efficiently index RDF data. We introduce compressed data structure to compactly represent RDF triples while guaranteeing fast pattern matching operations. The approach asks for 60% less space while speeding up query execution up to 81x. For time series, we developed a novel compression algorithm that achieves a significant improvement of compression ratio (up to 6.4x times) over the state-of-the-art method proposed by Facebook.


Innovation 11

Title: Deep Learning on Big Data from the winemaking industry

Description:  In BigDataGrapes, we developed machine learning techniques for the estimation of the number of leaves of the vine. The number of leaves is a key indicator of vegetative organogenesis in grapevine. The solution exploits state-of-the-art deep learning solutions based on a combination of Convolutional and Feedforward neural networks. The solution, trained on 250,000 photos, shows a mean average error of 0.87 (less that a single leaf per plant), making feasible a laborious and time-consuming task.