Agritech, autonomous growing systems case

Intro

AgwaFarm develops fully autonomous growing systems that bring fresh, nutrient-rich, chemical-free greens to maritime vessels and cruise liners. On long voyages, access to fresh produce is limited due to supply chain disruptions, rapid deterioration of stored foods, and the need for significant resources to maintain inventory. AgwaFarm eliminates this dependency by enabling crews to grow premium greens directly onboard - reliably, sustainably, and year round.

To achieve this, the company designed compact cultivation modules that maintain a self-regulated micro-climate with automatic control of lighting, irrigation, airflow, and nutrient delivery. At the heart of the system is the Virtual Agronomist, AgwaFarm’s AI-driven engine that continuously interprets sensor readings, monitors plant behavior, and adjusts environmental conditions to optimize growth and plant health. The system detects early signs of stress, tracks development patterns, and tailors growing plans to ensure consistent harvests with minimal human involvement.

This fully autonomous approach places exceptionally high requirements on the quality of visual data the AI relies on. Precise interpretation of plant stages, health cues, and dense growth patterns is essential for the Virtual Agronomist to make correct, real-time decisions, a need that ultimately led AgwaFarm to collaborate with Keymakr.

The challenge

As part of the project, the Keymakr team performed object detection, marking plants in the pots using bounding boxes. Additionally, through an attribute based system, the team classified plant growth stages to capture the developmental dynamics at each stage.

Accomplishing these tasks came with several nontrivial challenges. The plants inside the modules grow very close to one another, often intertwining or overlapping neighboring sprouts. Drawing a clear visual boundary between leaves can be difficult even for a trained expert, making accuracy harder to achieve.

Another challenge was growth-stage classification. Early signs of germination can be almost invisible in images: what appears to be an empty pot - or a new pod - may actually contain the very first millimeters of new sprouts. Mature plants, on the other hand, can fully cover younger ones, making their developmental stage hard to interpret visually.

As a result of this degree of ambiguity, the project required intensive communication, frequent clarifications, agronomist comments, and quick adjustments.

Results

Thanks to precise manual annotation, a flexible approach, and continuous communication, Keymakr helped AgwaFarm significantly improve the performance of their AI Virtual Agronomist. The collaboration delivered measurable outcomes that directly enhanced the efficiency of growing greens on marine vessels.

Model accuracy improvement: the Virtual Agronomist’s accuracy increased from 82% to 95%, significantly enhancing growth-stage recognition.

Flawless communication: an efficient feedback loop enabled quick clarification of complex cases and rapid corrections.

Flexibility as a competitive advantage: Keymakr’s ability to work efficiently with boutique, low-volume projects became a key advantage, providing the adaptability and precision required for this format.

“At Agwa, accurate vision is more than a model metric — it’s how we understand every plant, every day. Keymakr’s annotation expertise has helped us steadily sharpen our core detection model’s ability to recognize growth stages, planting events, harvest readiness, and early signs of plant stress. This ensures we can provide our customers with reliable, real-time insights that drive healthier crops and more efficient operations.

We deeply value Keymakr’s excellent collaboration, proactive communication, and careful attention to every small detail. Their commitment to quality has meaningfully strengthened the capabilities of our virtual agronomist — enabling it to make smarter decisions and support growers around the world.”

Rinat Landman, Data Scientist at Agwafarm

“This project is significant for us because we got to see how the quality of annotation directly affects the final product. The communication with AgwaFarm was incredibly professional - they were invested, shared context, demonstrated real growing modules, and helped us understand every detail. This type of partnership is genuinely inspiring, and I hope to one day try some of the fresh greens grown by this awesome system.”

Roman Gron, Keymakr PM

🍪 We use third party cookies to personalize content, ads and analyze site traffic. Learn more

Reviews
on

"Delivering Quality and Excellence"

The upside of working with Keymakr is their strategy to annotations. You are given a sample of work to correct before they begin on the big batches. This saves all parties time and...

"Great service, fair price"

Ability to accommodate different and not consistent workflows.
Ability to scale up as well as scale down.
All the data was in the custom format that...

"Awesome Labeling for ML"

I have worked with Keymakr for about 2 years on several segmentation tasks.
They always provide excellent edge alignment, consistency, and speed...

How Keymakr helped AgwaFarm train a Virtual Agronomist for on-board greens production on marine vessels