Image
Overview
This service consists of deploying a modular sensor payload (equipped with Lidar, inertial motion units, GNSS, Radar, stereo cameras, and ultra-wideband technology) to agricultural environments such as vineyards and orchards to collect and annotate high-fidelity datasets tailored to customer needs. Data are systematically labelled and can be delivered to agrifoodTEF customers as structured datasets for AI-driven agricultural solutions.
More about the service
Discover more about our service, including how it can benefit you, the delivery process, and the options for customisation tailored to your specific needs!
Problem (Before):
AgrifoodTEF customers (e.g., agricultural robotics/AI developers, precision farming companies, or researchers) often lack access to high-quality, annotated, real-world datasets from dynamic agricultural environments (e.g., vineyards, orchards).
Without this, they struggle to:
- Build robust AI systems and use cases (e.g., for crop monitoring, robotic harvesting, or autonomous navigation).
- Validate system performance under realistic, variable conditions (e.g., uneven terrain, changing light/weather).
- Benchmark their solutions against industry standards or competitors.
Solution (After): By deploying the sensor payload to collect and label data in real-world settings, the service enables customers to:
1. Build AI models effectively with diverse, context-rich datasets (e.g., labelled images of fruit clusters, 3D terrain maps).
2. Validate system reliability by testing solutions against benchmarks (e.g., "After using the service, a robotic harvester’s error rate in detecting ripe fruit drops from 25% to 8% in cluttered orchard conditions").
3. Identify and fix weaknesses (e.g., a navigation algorithm fails in GNSS-denied rows; the service provides UWB/IMU data to refine localisation).
4. Compare performance against competitors using standardised metrics (e.g., "Our weeding robot achieves 95% accuracy vs. the industry average of 82%").
This service can be delivered as an initial step prior to other services.
AgrifoodTEF customers (e.g., agricultural robotics/AI developers, precision farming companies, or researchers) often lack access to high-quality, annotated, real-world datasets from dynamic agricultural environments (e.g., vineyards, orchards).
Without this, they struggle to:
- Build robust AI systems and use cases (e.g., for crop monitoring, robotic harvesting, or autonomous navigation).
- Validate system performance under realistic, variable conditions (e.g., uneven terrain, changing light/weather).
- Benchmark their solutions against industry standards or competitors.
Solution (After): By deploying the sensor payload to collect and label data in real-world settings, the service enables customers to:
1. Build AI models effectively with diverse, context-rich datasets (e.g., labelled images of fruit clusters, 3D terrain maps).
2. Validate system reliability by testing solutions against benchmarks (e.g., "After using the service, a robotic harvester’s error rate in detecting ripe fruit drops from 25% to 8% in cluttered orchard conditions").
3. Identify and fix weaknesses (e.g., a navigation algorithm fails in GNSS-denied rows; the service provides UWB/IMU data to refine localisation).
4. Compare performance against competitors using standardised metrics (e.g., "Our weeding robot achieves 95% accuracy vs. the industry average of 82%").
This service can be delivered as an initial step prior to other services.
Service Delivery & Logistics
1. How is the service delivered?
Deployment Model:- On-site execution:
A sensor-equipped payload (modular, vehicle/robot-mounted) is deployed to the customer’s agricultural site (e.g., vineyard, orchard) or partner test farms.
- Repetitions: Data collection can be performed as a single campaign or recurring sessions (e.g., seasonal cycles, growth stages) to capture variability (e.g., flowering vs. harvest periods).
2. When can the service be delivered?
- Scheduling: Requires advance booking (e.g., 4–8 weeks) for payload deployment and site preparation.
3. How long does execution take?
- On-site deployment: 1–2 weeks per campaign (adjustable based on farm size and data density).
- Post-processing: Additional 2–3 weeks for labelling and report generation.
- Total timeline: ~4–8 weeks from deployment to delivery.
Location & Customer Requirements
4. Where is the service executed? - Customer’s own site: Farms, orchards, or vineyards.
- Partner test farms: Pre-approved agrifoodTEF satellite nodes (e.g., standardised test fields for benchmarking).
5. Customer location requirements:
- Regional focus: Prioritises regions with agrifoodTEF infrastructure (e.g., EU initially).
- Global access: Possible via partnerships with local nodes (additional logistics fees may apply).
Customer Deliverables
6. What does the customer receive?
Outputs:- Raw & labelled datasets: Structured, timestamped sensor data (Lidar point clouds, stereo camera images, GNSS/IMU trajectories) in AI-ready formats (e.g., ROS bags, COCO JSON).
- Documentation: Metadata (sensor specs, environmental conditions), annotation guidelines, and compliance certificates.
- Summary report: Actionable insights (e.g., “System struggles in low-light conditions; recommend augmenting training with twilight radar data”).
Customer Obligations
7. What must the customer provide?
- Site access: Safe, operational fields for deployment (e.g., cleared rows for robot/sensor mobility).
- Use case specifications:
- Target metrics (e.g., “Collect data on apple cluster density in rows 5–10”).
- AI/system requirements (e.g., “Annotate fruit ripeness labels for YOLO training”).
- Ethical/legal compliance: Permissions for data collection (if required).
1. How is the service delivered?
Deployment Model:- On-site execution:
A sensor-equipped payload (modular, vehicle/robot-mounted) is deployed to the customer’s agricultural site (e.g., vineyard, orchard) or partner test farms.
- Repetitions: Data collection can be performed as a single campaign or recurring sessions (e.g., seasonal cycles, growth stages) to capture variability (e.g., flowering vs. harvest periods).
2. When can the service be delivered?
- Scheduling: Requires advance booking (e.g., 4–8 weeks) for payload deployment and site preparation.
3. How long does execution take?
- On-site deployment: 1–2 weeks per campaign (adjustable based on farm size and data density).
- Post-processing: Additional 2–3 weeks for labelling and report generation.
- Total timeline: ~4–8 weeks from deployment to delivery.
Location & Customer Requirements
4. Where is the service executed? - Customer’s own site: Farms, orchards, or vineyards.
- Partner test farms: Pre-approved agrifoodTEF satellite nodes (e.g., standardised test fields for benchmarking).
5. Customer location requirements:
- Regional focus: Prioritises regions with agrifoodTEF infrastructure (e.g., EU initially).
- Global access: Possible via partnerships with local nodes (additional logistics fees may apply).
Customer Deliverables
6. What does the customer receive?
Outputs:- Raw & labelled datasets: Structured, timestamped sensor data (Lidar point clouds, stereo camera images, GNSS/IMU trajectories) in AI-ready formats (e.g., ROS bags, COCO JSON).
- Documentation: Metadata (sensor specs, environmental conditions), annotation guidelines, and compliance certificates.
- Summary report: Actionable insights (e.g., “System struggles in low-light conditions; recommend augmenting training with twilight radar data”).
Customer Obligations
7. What must the customer provide?
- Site access: Safe, operational fields for deployment (e.g., cleared rows for robot/sensor mobility).
- Use case specifications:
- Target metrics (e.g., “Collect data on apple cluster density in rows 5–10”).
- AI/system requirements (e.g., “Annotate fruit ripeness labels for YOLO training”).
- Ethical/legal compliance: Permissions for data collection (if required).
Sensor Configuration:
- Customers can select specific sensors from a predefined list of candidate sensors for the payload (e.g., prioritise Lidar + stereo cameras for 3D mapping, disable GNSS for indoor testing).
Data Annotation:
- Choose annotation formats from a predefined list (e.g., bounding boxes, semantic segmentation masks) compatible with their AI pipelines.
Temporal/Seasonal Flexibility:
- Schedule recurring campaigns (e.g., weekly data collection during harvest season, multi-year phenological studies).
Seasonal restrictions:
- Aligns with vegetation periods (e.g., no frost/rain for certain sensors; optimal timing for crop-specific data like fruit maturity in orchards).
- Example: Vineyard data collection restricted to the growing season (April–October in EU climates).
What Customers Should Know in Advance
- Site Preparation: Customers must provide field maps, irrigation schedules, and safety protocols.
- Data Ownership: Raw data is shared with the customer, but agrifoodTEF may use aggregated, anonymised data for internal R&D.
- Cost Implications:
- Recurring campaigns or ultra-high-resolution sensors incur higher fees.
- Post-processing of complex annotations (e.g., 3D point cloud labelling) billed separately
- Customers can select specific sensors from a predefined list of candidate sensors for the payload (e.g., prioritise Lidar + stereo cameras for 3D mapping, disable GNSS for indoor testing).
Data Annotation:
- Choose annotation formats from a predefined list (e.g., bounding boxes, semantic segmentation masks) compatible with their AI pipelines.
Temporal/Seasonal Flexibility:
- Schedule recurring campaigns (e.g., weekly data collection during harvest season, multi-year phenological studies).
Seasonal restrictions:
- Aligns with vegetation periods (e.g., no frost/rain for certain sensors; optimal timing for crop-specific data like fruit maturity in orchards).
- Example: Vineyard data collection restricted to the growing season (April–October in EU climates).
What Customers Should Know in Advance
- Site Preparation: Customers must provide field maps, irrigation schedules, and safety protocols.
- Data Ownership: Raw data is shared with the customer, but agrifoodTEF may use aggregated, anonymised data for internal R&D.
- Cost Implications:
- Recurring campaigns or ultra-high-resolution sensors incur higher fees.
- Post-processing of complex annotations (e.g., 3D point cloud labelling) billed separately
Location
At user's premises
Italy
Type of Sector
Arable farming
Greenhouse
Horticulture
Tree Crops
Viticulture
Type of service
Collection of test data
Accepted type of products
Physical system
Software or AI model
