Catalogue of Services

Provision of datasets

Politecnico di Milano (POLIMI)

Università degli Studi di Milano (UMIL)

Location

Italy

Remote

Arable farming

Food processing

Greenhouse

Horticulture

Livestock farming

Tree Crops

Viticulture

During the development of systems based on AI and/or robotics, tailored data are needed to support the realisation, validation and optimisation of systems and subsystems. This service aims at providing such data.

AI models are a noteworthy example of systems that require data to be trained; more generally, any system which processes data produced via artificial perception (such as images, videos, LiDAR, ultrasound, radar...) requires suitable data to be built, tested and fine-tuned.

This service allows a customer to define, together with agrifoodTEF, the specific features of the data that the customer’s activities require and to receive datasets precisely matching such features. Among the features to be defined are the type, number and specifications of the data streams comprised in the datasets (e.g., when multi-sensor data are required) and the range of variations encompassed by the data (e.g., the variety of environmental conditions represented in a sensor stream); lower-level features to be defined concern the formatting and packaging of the data.

Support in interconnecting systems to AgrifoodTEF’s computational test infrastructure

Politecnico di Milano (POLIMI)

Università degli Studi di Milano (UMIL)

Location

Italy

Remote

Arable farming

Food processing

Greenhouse

Horticulture

Livestock farming

Tree Crops

Viticulture

Digital testing requires interaction among several different components which have to reliably cooperate to enable the testing activities and to ensure that their outcomes reliably reflect the system’s capabilities. When digital testing of customer systems leverages agrifoodTEF’s computational infrastructure, integration between the infrastructure and the systems to be tested is needed before the tests/experiments can be executed. If the customer does not possess the technical expertise – or does not want to devote resources – to perform the integration, this service can support them in the process. Help is provided by a team of expert engineers.

This service concerns the integration of the customers’ system and/or data used for testing with the agrifoodTEF digital testing infrastructure. In particular, integration is crucial to ensure smooth and uninterrupted communication between the system and the infrastructure that produces, collects and/or processes test data.

This service also includes the development of any custom components needed to ensure a two-way integration between the system and the reference infrastructure (e.g., software components to format data or to perform transcoding).

Support in interconnecting systems to physical infrastructure

Politecnico di Milano (POLIMI)

Università degli Studi di Milano (UMIL)

Location

Italy

Arable farming

Greenhouse

Horticulture

Tree Crops

Viticulture

Testing a system in a physical environment requires an environment that has been prepared and set up to support the specific testing activities to be executed. Such an environment usually involves both agronomy-related elements (machines, fields, greenhouses, cultivations...) and technical infrastructure (sensors, networks, software, logging systems...). In order to execute the testing, the physical system under test (e.g., an agricultural robot) must correctly interface with all these elements; this service takes care of ensuring this.

For instance, it is crucial to ensure that the communication between the system under test and the technical infrastructure supporting the testing activities is correct, smooth and uninterrupted.

Wherever needed, the service also includes the development of custom components needed to ensure a successful two-way integration between the system and the testing infrastructure (e.g., a software component to format data or perform transcoding).

While available separately, this service is often used in association with service S00110 (Preparation of physical test environment).

Testing of AI-based sensor performance

Josephinum Research (JR)

Location

At user's premises

Austria

Arable farming

Horticulture

Tree Crops

Viticulture

Our service thoroughly evaluates AI-based sensors by measuring key metrics like accuracy, mean squared error, and other tailored parameters. Testing is conducted in real-world environments where the sensor will be used, with reference data recorded or labelled for analysis. For certain sensors, we also utilise benchmark datasets to validate performance. We assess consistency under identical conditions, adaptability to environmental changes (e.g., light, temperature, or weather), and improvements in accuracy over time. Additionally, we measure power consumption, processing efficiency, and memory usage to ensure optimal resource utilisation. Response time between detection and action is evaluated for real-time applications, while extreme condition testing (e.g., bright light, rainy conditions, compacted soil) ensures robustness in challenging environments. We also verify data security and resilience against adversarial attacks. This comprehensive testing is essential, ensuring sensors meet high standards of precision, reliability, and efficiency. Benchmark datasets, standardised data collections used for validation, provide a reliable baseline for performance comparison. By addressing these aspects, our service ensures your sensor technology is ready for real-world deployment, delivering the performance and adaptability your application demands.

Testing of autonomous carrier & implement integration

Josephinum Research (JR)

Location

At user's premises

Austria

Arable farming

Horticulture

Tree Crops

Viticulture

Evaluating an agricultural implement on an autonomous carrier platform involves more than just numerical metrics; it includes a qualitative analysis of various aspects. We assess work quality by examining performance under different agricultural conditions, including, among others, different soil moisture levels, varying crop densities, and sloped terrain. Safety evaluations consider system robustness in various scenarios. Functionality assessments measure adaptability in dynamic environments, accounting for sudden weather changes and unexpected challenges.

Testing of biomass and yield estimation

Josephinum Research (JR)

Location

At user's premises

Austria

Remote

Arable farming

Our testing service assesses the accuracy of your system's biomass estimation capabilities. You can opt to evaluate your model virtually using existing datasets, such as satellite or image data, or we can gather new field data with our field sampling equipment specifically for your needs. The main targeted crops of the biomass estimation are grassland as well as clover and alfalfa, but also grain crops are possible. As yield and biomass are highly correlated in grassland, both can be determined with this test. We provide results using metrics like Pearson correlation/R-squared and MSE/MAE to give you a clear picture of your model's performance.

Testing of field navigation

Josephinum Research (JR)

Location

At user's premises

Austria

Arable farming

Horticulture

Tree Crops

Viticulture

Our service evaluates the navigation accuracy and performance of autonomous systems directly in the field. We conduct comprehensive tests under real-world conditions to ensure that your system navigates precisely and efficiently. This testing is crucial for startups and SMEs developing innovative solutions that require dependable autonomous navigation capabilities.

Testing of plant detection

Josephinum Research (JR)

Location

Remote

Arable farming

Horticulture

Tree Crops

Our digital test helps you evaluate how well your model identifies different types of plants. Your model can be delivered in a docker container or the inference results are collected in real time from your device in the field. We focus on key performance metrics like accuracy, precision, recall, Intersection over Union (IoU), and F1 score to give you a clear picture of your model's effectiveness. Crucial to this process is the ground truth data, which can be either specially prepared for this test or sourced from an existing benchmark dataset.

Testing of spot spraying performance

Josephinum Research (JR)

Location

At user's premises

Austria

Arable farming

Horticulture

Our spot spraying test service precisely checks how well your spraying system works under real-world conditions. The innovative approach allows for immediate documentation of the spray distribution without altering the spraying process. After application, we document the spread pattern and analyse this data to assess how effectively and precisely the liquid targets the intended areas. This enables you to fine-tune application techniques and evaluate herbicide effects, ensuring your spraying strategies are both efficient and environmentally conscious.

Testing of the follow-me function of a robot

Josephinum Research (JR)

Location

Austria

Arable farming

Horticulture

Livestock farming

Tree Crops

Viticulture

Our service offers a comprehensive evaluation of your device's ability to detect and track moving objects like people, ensuring it performs reliably in real-world scenarios. Also, safety concerns of such a function can be addressed with our tests. Through field tests, we simulate realistic conditions to test your system under various environmental influences. This practical approach provides precise insights into your device's effectiveness and highlights areas for improvement, making it essential for startups and SMEs in the tech-agricultural sector aiming to refine their innovative solutions.

Testing of the person detection function of a robot

Josephinum Research (JR)

Location

At user's premises

Austria

Arable farming

Horticulture

Tree Crops

Viticulture

This service tests the robustness and accuracy of a solution to detect moving objects, such as people. We conduct these tests manually in the field to ensure that your device is evaluated under realistic conditions, providing you with reliable insights into its performance and areas for improvement. The test will be done under various conditions (different illuminations, rain, sunshine, diffuse light, etc.) to address the needs of your customers. You will receive a report of the system's capabilities as well as under which conditions the detection should be improved.

Testing of weeding performance

Josephinum Research (JR)

Location

At user's premises

Austria

Arable farming

Horticulture

Our service offers comprehensive field trials to test the performance of your AI-based weeding system. Before the system operates, we record essential environmental data such as soil conditions and weather. We assess the crops and weeds either manually or with automated tools, and re-evaluate them post-weeding to measure the system's effectiveness. Our trials can be conducted on fields provided by us or at external locations of your choosing. This testing gives you clear, objective insights into your system’s performance and helps identify any potential issues under real-world conditions.