1. Data Science To Tackle Innovation Barriers To
Produce Healthy Sustainable Food
By: Jakob de Vlieg
Global Challenge To Meet Food Supply
● Reasons: expanding population, shortage of agriculture land, water, and energy, and
climate change.
● Same time: the need for increased food production, and lots of food production is
wasted.
FAIR-Data
● Findable: (meta) data uniquely and persistently identifiable.
● Accessible: identifiers should provide a mechanism for (meta) data access, incl.
authentication, access protocols, license, etc.
● Interoperable: (meta) data should be machine-readable and annotated with
resolvable vocabularies/ontologies.
● Reusable: (meta) data is sufficiently well described to allow (semi) automated
integration with other compatible data sources.
▶ AI allows food value chain integration and chain reversal.
Europe Ambition
Reducing the use of mineral fertilizer by 30% and chemical pesticides by 50% → requires an
effective digitalization strategy to implement vision and goals.
Precision Farming Dependent on Data Fusion, Interoperability, and Security
Reinforcement Learning
No guidance or human expertise is needed. Learn by running millions of tests and giving
rewards for reaching certain goals.
1
,Data-Centric AI
● Shift focus on AI/model algorithm development towards getting high-quality labeled
data.
● Develop a new systematic mental framework.
State Agriculture and Food Sector
● High-tech and AI have a huge impact on the future of the agriculture and food sector.
● Increasing labor costs and shortages in the agricultural and food sector.
● Increasing investment in the agriculture and food technology sector.
Legislative Bottlenecks
Legislation in the EU does not allow the autonomous use of field robots in open cultivation.
● Allowed in: Australia, USA, and Japan.
▶ Systematic transformation is needed at unprecedented speed and scale, and requires
other ingredients:
● Evolution of public consciousness.
● Social movements.
● Good storytelling.
● Money.
● Political pressure.
● Visionary leaders.
Data & AI Opportunities To Reduce Food Waste
● Blockchain-enabled traceability.
● IoT for real-time supply chain transparency and traceability.
● Big data & advanced analytics for insurance.
Short Chain Concept
Produce local, unique food products delivered to your doorstep:
● Miniaturized food processing equipment.
● Sensing & data technology to validate the quality and safety of locally produced food
products remotely.
● (Last mile) logistics for fresh food products.
Data Fusion
Microservice and layered architecture to deal with many (distributed) cross-type data
sources and sensors.
Mixed Cropping
Growing more than one crop on the same field.
● Crop-to-crop interactions to protect against pets.
● Better for the soil.
● Higher biodiversity.
▶ Mixed cropping is very labor intensive → agribots & AI could provide solutions.
2
, Vertical Farm
Plants grow in stacks near the consumers in a controlled environment. No pesticides need to
be used, no dependency on the season, less water used, and better use of land.
● Limited to a small number of crops.
● High startup and energy costs.
Self-supervised Learning
Unsupervised learning method where the supervised learning task is created using:
● Unlabelled data.
● Previous self-supervision approaches.
Models-as-a-service
Speed up product development in the food sector by data science allowing sharing of
proprietary food models.
● Reduce development costs.
● Change the way of working by joining forces.
● Speed up product development.
2. Sensor Measurements & Timeseries
By: Sicco Pier van Gosliga & Rogier Brussee
Accelerometer
Measures acceleration on each axis in g-force (9.8 m/s2).
● Measures 3-axis + temperature.
Measurements
Tell us about a property of something and give a number to that property.
● Counting is (most of the time) not really a measurement.
● Experiment or test is not really a measurement.
Error & Uncertainty
Uncertainty of measurement is doubt that exists about the result of any measurement.
● Error: the difference between the measured value and the true value (expressed as
an interval).
● Uncertainty: quantification of the doubt about the measurement result (expressed as
confidence level).
Arithmetic Mean
Also known as the average, is shown as the x̄ (x-bar) symbol, the mean value of x.
Standard Deviation
Repeated measurements give different results, the spread between the measurement is
known as the standard deviation.
3
Produce Healthy Sustainable Food
By: Jakob de Vlieg
Global Challenge To Meet Food Supply
● Reasons: expanding population, shortage of agriculture land, water, and energy, and
climate change.
● Same time: the need for increased food production, and lots of food production is
wasted.
FAIR-Data
● Findable: (meta) data uniquely and persistently identifiable.
● Accessible: identifiers should provide a mechanism for (meta) data access, incl.
authentication, access protocols, license, etc.
● Interoperable: (meta) data should be machine-readable and annotated with
resolvable vocabularies/ontologies.
● Reusable: (meta) data is sufficiently well described to allow (semi) automated
integration with other compatible data sources.
▶ AI allows food value chain integration and chain reversal.
Europe Ambition
Reducing the use of mineral fertilizer by 30% and chemical pesticides by 50% → requires an
effective digitalization strategy to implement vision and goals.
Precision Farming Dependent on Data Fusion, Interoperability, and Security
Reinforcement Learning
No guidance or human expertise is needed. Learn by running millions of tests and giving
rewards for reaching certain goals.
1
,Data-Centric AI
● Shift focus on AI/model algorithm development towards getting high-quality labeled
data.
● Develop a new systematic mental framework.
State Agriculture and Food Sector
● High-tech and AI have a huge impact on the future of the agriculture and food sector.
● Increasing labor costs and shortages in the agricultural and food sector.
● Increasing investment in the agriculture and food technology sector.
Legislative Bottlenecks
Legislation in the EU does not allow the autonomous use of field robots in open cultivation.
● Allowed in: Australia, USA, and Japan.
▶ Systematic transformation is needed at unprecedented speed and scale, and requires
other ingredients:
● Evolution of public consciousness.
● Social movements.
● Good storytelling.
● Money.
● Political pressure.
● Visionary leaders.
Data & AI Opportunities To Reduce Food Waste
● Blockchain-enabled traceability.
● IoT for real-time supply chain transparency and traceability.
● Big data & advanced analytics for insurance.
Short Chain Concept
Produce local, unique food products delivered to your doorstep:
● Miniaturized food processing equipment.
● Sensing & data technology to validate the quality and safety of locally produced food
products remotely.
● (Last mile) logistics for fresh food products.
Data Fusion
Microservice and layered architecture to deal with many (distributed) cross-type data
sources and sensors.
Mixed Cropping
Growing more than one crop on the same field.
● Crop-to-crop interactions to protect against pets.
● Better for the soil.
● Higher biodiversity.
▶ Mixed cropping is very labor intensive → agribots & AI could provide solutions.
2
, Vertical Farm
Plants grow in stacks near the consumers in a controlled environment. No pesticides need to
be used, no dependency on the season, less water used, and better use of land.
● Limited to a small number of crops.
● High startup and energy costs.
Self-supervised Learning
Unsupervised learning method where the supervised learning task is created using:
● Unlabelled data.
● Previous self-supervision approaches.
Models-as-a-service
Speed up product development in the food sector by data science allowing sharing of
proprietary food models.
● Reduce development costs.
● Change the way of working by joining forces.
● Speed up product development.
2. Sensor Measurements & Timeseries
By: Sicco Pier van Gosliga & Rogier Brussee
Accelerometer
Measures acceleration on each axis in g-force (9.8 m/s2).
● Measures 3-axis + temperature.
Measurements
Tell us about a property of something and give a number to that property.
● Counting is (most of the time) not really a measurement.
● Experiment or test is not really a measurement.
Error & Uncertainty
Uncertainty of measurement is doubt that exists about the result of any measurement.
● Error: the difference between the measured value and the true value (expressed as
an interval).
● Uncertainty: quantification of the doubt about the measurement result (expressed as
confidence level).
Arithmetic Mean
Also known as the average, is shown as the x̄ (x-bar) symbol, the mean value of x.
Standard Deviation
Repeated measurements give different results, the spread between the measurement is
known as the standard deviation.
3
