Petiole (2026)

15/03/2026

Accurate measurement matters most when work moves from a single sample to a batch.

One leaf measured badly is a mistake.
A whole batch measured badly becomes a false conclusion.

In plant phenotyping, quality control, and crop monitoring, batches are where decisions start to scale. A single image can look convincing, but batch analysis reveals whether a method is truly reliable. Are measurements consistent across many leaves, many plants, many trays, many days? Can the workflow handle natural biological variation without drifting? Can it separate real differences from noise?

That is why accuracy is not only about getting a nice number. It is about building trust in the whole dataset.

When measurements are accurate across a batch, we can:
✅compare samples with confidence
✅detect meaningful differences earlier
✅reduce subjective bias
✅support more reliable reporting
✅make better operational and research decisions

This becomes especially important in agriculture, where decisions are often made on groups rather than individuals: a tray, a greenhouse zone, a field block, a genotype set, a treatment batch. If batch measurements are unstable, the whole interpretation becomes fragile. But if they are robust, even simple metrics such as leaf area become powerful.

Good batch measurement turns images into evidence.

And evidence is what allows automation to move from “interesting” to genuinely useful.

At Petiole Pro, this is one of the principles behind our work: not just measuring plants, but measuring them in a way that remains dependable when the sample size grows.

For leaf area analysis, batch accuracy means one key thing: the numbers should remain trustworthy not only for one leaf, but for hundreds.

How many leaves, seeds, or fruits do you need to measure? Tell us your batch size, and we’ll recommend the most suitable data capture protocol.

29/12/2025

At Petiole Pro during 2025...

- We’ve done real work for real people.
- We’re grateful for you and delighted that you've chosen us.
- We’re still growing and have had some interesting ideas to bring into the reality.

Thank you for being part of the journey.
Come with us into 2026 🌱🌍🚀

11/12/2025

Before you see our new mobile application, we’d like to give you some background on how it works.
The most important part of our new app is the ontology.

Ontology is the unglamorous backbone of agrifood productivity.

It’s the shared language that tells humans, sensors, and software what a “lot,” “defect,” “plot,” “task”, or “ripeness” actually means — so data can move cleanly from field to lab to packhouse to ERP without being mistranslated on every hop.

Why it matters for productivity and automation:

1. 𝗖𝗹𝗮𝗿𝗶𝘁𝘆 → 𝗳𝗲𝘄𝗲𝗿 𝗲𝗿𝗿𝗼𝗿𝘀. When “Brix,” “size class,” and “moisture” are defined once and reused everywhere, teams stop arguing about columns and start acting on signals. Decisions speed up.

2. 𝗜𝗻𝘁𝗲𝗿𝗼𝗽𝗲𝗿𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗯𝘆 𝗱𝗲𝘀𝗶𝗴𝗻. A good ontology maps drones, weather APIs, lab instruments, and QC apps onto the same concepts. That means plug-and-play integrations instead of brittle one-off scripts.

3. 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻 𝘁𝗵𝗮𝘁 𝘀𝗰𝗮𝗹𝗲𝘀. Workflows like “harvest → intake QC → grading → release gate” need machine-readable definitions of objects (fruit, bin, pallet), properties (defects, weight, temperature), and actions (scan, weigh, sort). Ontology turns SOPs into code.

4. 𝗕𝗲𝘁𝘁𝗲𝗿 𝗔𝗜. Models don’t just need data; they need meaning. If your “defect” label drifts site to site, your detection model won’t generalize. A domain ontology stabilizes training targets and makes features reusable across crops and seasons.

5. 𝗧𝗿𝗮𝗰𝗲𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝘁𝗵𝗮𝘁’𝘀 𝗿𝗲𝗮𝗹-𝘁𝗶𝗺𝗲. When links between entities are explicit (block → plot → lot → pallet → shipment), audits move from weeks to minutes and root-cause analysis becomes a query, not a war room.

The payoff of clear ontology?
Fewer spreadsheets, faster cycles, trustworthy metrics — and models & agents that actually do work, not demos 😉

Ontology is not bureaucracy; it’s an operating system for efficient business.
Build it once, and every new tool, dashboard, and AI model gets easier — and more valuable.

Tomorrow we will shed more light about the ontology.
Stay tuned 🌱

10/12/2025

Ontology is, in simple terms, a common data language for agrifood production.

When we talk about digitalisation, this becomes essential:
-> devices,
-> field teams,
-> labs,
-> ERPs,
-> and AI models all need to “understand” the same things in the same way.

Without that shared vocabulary, data stays siloed and automation becomes fragile.

🌱🌱🌱

In the pre-harvest stage, an ontology should describe objects like farm, block, plot, crop, variety, growth stage, task, input (fertiliser, water, pesticide), sensor, and event (disease alert, weather anomaly). Each of these needs standard properties: dates, locations, units, responsible person, thresholds. If a scouting app says “early blight” and a decision-support tool says “disease level 2,” but they aren’t mapped to the same concept, the system can’t trigger the right action. A good pre-harvest ontology lets you connect agronomy, monitoring, and task management in one flow.

🌾🌾🌾

In the post-harvest stage, the ontology shifts to intake and quality. Now we talk about lot, batch, bin, pallet, product, defect, grade, temperature, photo evidence, and release criteria. Here the challenge is traceability and consistency: intake photos must link to the same lot ID that later appears in QC, packing, and dispatch.

If “Russeting_L1” in one system is “SurfaceDefectMinor” in another, analytics will be unreliable.
A post-harvest ontology standardises grades, defect names, measurement methods, and status transitions (received → inspected → approved/rejected).

Why does this matter?
1. Because digitalisation is not only about collecting data; it’s about making data interoperable.
2. Once pre- and post-harvest speak the same language, you can track performance from field to packhouse, train AI on clean labels, and automate approvals with confidence.
3. That’s why we keep saying: ontology is not theory. It’s the layer that makes agrifood automation possible.

How do you define ontology in your agrifood business?

09/12/2025

When we bring ontology into the topic of labour — pickers, seasonal workers, field teams, supervisors — we’re doing the same thing as with crops and lots: giving the system a precise, shared way to describe who is doing what, where, and for how long.

In a labour context, an ontology defines core entities such as:
-> Employee/Worker,
-> Role (picker, quality checker, supervisor),
-> Task (harvest row 12, sort Grade A, load pallet),
-> Location/Block,
-> Shift, and
-> Output (kg picked, crates filled, rows completed).

Each of these gets standard properties: start/end time, rate, crop, variety, team, device used, and sometimes even skill level.

When this is defined once, different modules — HR, field app, payroll, productivity dashboard — can all talk about labour in the same terms 🤝

Why is that important?
Because in many farms and packhouses, labour data is collected, but not aligned.
One person records “Team Maria,” another “Picker 14,” another just “3 workers 05.11”
Without an ontology, those can’t be reconciled, so you cannot answer simple management questions like: “Which team was most productive on Gala today?” or “How much labour per ton did block A actually require?”

👉 An ontology also helps connect labour → task → output.
For example:
Task: “Harvest blueberries, Block 5”
Assigned to: “Seasonal_picker”
Measured output: “Crate_125, 4.2 kg, 10:34”
Because all three use the same identifiers, the system can later calculate labour cost per kg, spot underperforming rows, or trigger extra pickers to a field that is behind plan.

Finally, once labour concepts are formalised, automation becomes possible: ✅ the system can auto-assign tasks to available roles,
✅ flag missing check-ins, or
✅ suggest optimal crew sizes based on historical ontology-backed data.

So even for people-related processes, ontology is the bridge between “we have records” and “we can manage in real time”

If you see the gap in the management and performance of farm labor - we will be glad to support you with our new mobile-based, offline-first and photo-centric tool (plus a bit of AI on device😉)

[email protected]

08/12/2025

Ontology for greenhouse phenotyping sounds abstract, but it’s actually one of the most practical things you can put in place if you want reliable AI, comparable trials, and clean reporting across seasons and teams.

In a greenhouse we measure a lot:
-> plant height,
-> leaf number,
-> flowering stage,
-> fruit set,
-> colour,
-> disease symptoms,
-> substrate EC,
-> microclimate,
-> even images from fixed cameras.

Different teams often name the same thing differently (“plant ID,” “line,” “tray,” “slot”), and different devices export different column names.
That’s exactly where ontology helps.

What is it here?
A shared, machine-readable vocabulary that describes:
1. the experiment (trial, treatment, cultivar, replication),
2. the plant unit (plant, pot, gutter position),
3. the phenotypic trait (height, NDVI, leaf area, fruit count),
4. the observation event (who measured, with what device, at what time),
5. and the environment (zone, bed, compartment, setpoint).

When these are standardised, three big things happen:

🥇 Data becomes comparable.
If “plant_height_cm” always means the same trait, with the same unit and method, you can compare week 3 vs week 7, or even Trial A vs Trial B from last year.

🥈 Vision/AI becomes reusable.
If your camera pipeline outputs “fruit_count” or “leaf_area” using the same ontology terms as your manual scoring sheet, you can mix human and machine observations in one dataset without endless cleaning.

🥉Automation becomes safe.
Growth deviations, stress signals, or poor-performing cultivars can trigger actions (alert, re-measure, change climate recipe) because the system knows exactly what the measurement represents.

Greenhouse teams are already digital, but often not semantic. Getting the ontology right is the difference between “we have files” and “we have a phenotyping platform.”

Do you have ontology embedded into your standard operational procedures? 🌱

07/12/2025

Seed quality assessment sounds very “lab-and-protocol,” but it also has an information problem: the same seeds are described differently by different labs, different seasons, and different systems.

That’s exactly where an ontology helps — it gives everyone (and every device) the same language for what was tested, how, and what the result means.

What is “ontology” here?
A structured vocabulary that defines:
🌱 the seed lot (origin, species, variety, production year),
🌱 the sample (subsample, weight, method of sampling),
🌱 the test type (germination, purity, vigour, moisture, health/pathogen),
🌱 the method/protocol (ISTA rule, lab SOP, equipment used),
🌱 and the result (value, unit, rating, pass/fail, tolerance).

Why does this matter?
Comparability.
If one lab says “germination 92%” and another says “final count 92%” but they used different days or substrates, those data shouldn’t be treated as the same. An ontology stores method + context, so numbers stay meaningful.

Traceability from field → lot → lab.
A seed lot tested in January must still be recognisable in June when it is treated, packed, and shipped. If the lot ID, crop name, and variety are defined once and reused, quality data can follow the lot through the whole chain.

Automation and alerts.
When test results are machine-readable (“moisture > 13%” or “pathogen detected: Fusarium”), systems can trigger actions: re-dry, re-test, block the lot, or notify QA. No human re-interpretation needed.

Better AI/vision downstream.
If you later add image-based seed counting, damage detection, or purity classification, models can write back results using the same ontology terms as the lab. That keeps datasets clean and improves model training.

Seed businesses talk a lot about “consistency”
Ontology is how you make consistency digital — so that quality is not only measured, but understood by every system in the chain. 🌾📊

What do you think about this topic?

06/12/2025

🌿 What's similar between stomata and ontology?
Just one simple example.
“Physiology / Gas-Exchange” module to think (and soon - use!) in Petiole Pro

In applied agritech, gas-exchange traits tell us how a plant breathes, cools, and manages water. These traits are early and sensitive indicators of stress.

They are also measurable through sensors, imaging and lightweight AI models. That makes them perfect for automation.
A practical ontology follows a simple chain:

1. Seedling → Leaf → Physiological Traits → Gas-Exchange Traits → Stomatal Traits

2. Under this, we map the core variables:
• Stomatal density
• Stomatal aperture
• Stomatal conductance
• Transpiration efficiency
• Water-use efficiency (WUE)
• Leaf greenness / DGCI

3. Each trait is defined by what it measures, how it is measured, units, expected ranges, and interpretation rules.

Stomata count becomes “Stomata per ROI” or “Stomatal Density”, linked to leaf surface, crop species and developmental stage. DGCI connects colour to physiological state.

🌱 Conductance links to vapour exchange.
🌱 WUE connects physiology to agronomy.
🌱 The value of this ontology is consistency.

AI systems can grade seedlings, detect stress, and flag anomalies across nurseries, seasons and devices.
Agronomists can compare traits without ambiguity.

Data becomes interoperable, reusable, and verifiable.
This is how physiology becomes computable.
And how plants start speaking a language AI can understand.

Which other use cases of AI for stomata do you have in your mind?

23/11/2025

🌱 What does “quality” really mean for a cabbage seedling?
And who said the word "ontology" again???
🤖

Briefly, QA includes:
(1) data collection,
(2) analysis (one tiny bit is in the image)
(3) and, finally, reporting

Ontology is the backbone of these three parts.
An ontology for cabbage seedling QA gives a structured, shared language for evaluating what “good” looks like.

It links morphology, vigour, health, environment and actions

As a result, agronomists, nurseries and AI systems all speak the same language.

At the core is the Seedling object, branching into morphology (hypocotyl length, stem thickness, cotyledon symmetry, first-leaf size, early rosette shape), vigour indicators (greenness index, compactness, turgor, growth rate), and root-system descriptors.

Plus, cabbage especially rewards compactness—leggy seedlings rarely survive field life.
This could be also included in QC assessment screen 🫡

Health traits map visible symptoms (chlorosis, purpling, edge burn, damping-off, lesion patterns) to likely causes: nutrient imbalance, heat/light stress, overwatering, fungal pathogens or pests.

Each symptom comes with severity and localisation descriptors, making it usable for both human scorers and computer vision.

Environmental context such as substrate, EC, pH, irrigation, temperature, humidity, light quality—is tied to expected trait ranges and risk classes. The ontology then supports QA tasks: - grading,
- sorting,
- anomaly detection,
- disease-onset flags,
- and generating simple corrective recommendations.

Finally, digital phenotyping layers in masks, leaf contours, rosette symmetry keypoints and compactness ratios, enabling consistent QC across nurseries, seasons and automation setups.

How are you checking all these features of cabbage seedlings now?

12/10/2025

In computer vision, the most proven, ROI-positive use isn’t sci-fi—it’s quality control on the production line. Either on a conveyor belt or across the grading table, CV earns its keep by making calls that are fast, repeatable, and defensible.

Why? Because when a standard operating procedure (SOP) is followed, the environment is stable. Lighting, lens, distance, background—locked. The only variable is the object: berries, fruits, vegetables, tubers, nuts, even packaging units like clamshells, lids, labels, trays, cartons, and pallets.

👉 That stability turns images into decisions.

We talk with blueberry growers a lot. Nine times out of ten, two levers define post-harvest QC:

Size

Absence of visual defects

They’re right. In practice, that means monitoring calibre consistency (yes, a uniformity index matters), and detecting misshapes, scars, stem-end tearing, shrivel, splits, mould, and foreign matter. On the line you also watch for clamshell under-fill—a classic vision task.

Here’s the under-used advantage: you don’t need to invent your own rulebook. Blueberries have solid global standards:

🔵 UNECE FFV-57 for fresh: minimum quality, tolerances, packing.

🔵 OECD Berry Fruits: photo examples so inspectors interpret defects consistently.

🔵 Shipping to the U.S.? Align with USDA blueberry grades.

Working IQF? Anchor specs to Codex CXS 103 for quick-frozen blueberries.

🔵 For auditability end-to-end: GLOBALG.A.P. on farm + BRCGS/SQF in the packhouse.

Why this matters:

1. Clear standards → clear labels.
2. Clear labels → cleaner training data & fewer intake disputes.
3. Net result: fewer claim disagreements, faster lot release, tighter spec adherence, and dashboards that actually mean something to sales and growers.

Computer vision should count, measure, and flag. People make the judgement calls and press “Send report.” That’s how CV really works today—practical, defensible, and scalable.

What’s your experience implementing computer vision for QC?

16/07/2025

🟡 𝘿𝙧𝙤𝙨𝙤𝙥𝙝𝙞𝙡𝙖 𝙨𝙪𝙯𝙪𝙠𝙞𝙞 𝘼𝙬𝙖𝙧𝙚𝙣𝙚𝙨𝙨 𝘿𝙖𝙮

Spotted-wing drosophila (SWD) remains one of the biggest headaches for fruit growers — but smarter monitoring is changing the game.

Automated counts from yellow sticky traps using a mobile application turn hours of manual scouting into quick, data-rich insights that boost orchard productivity:

→ Timely thresholds – Instant counts reveal when pest numbers hit action levels, so interventions arrive on time.

→ Optimised inputs – Spray only the blocks that need it, cutting pesticide use and protecting beneficial insects.

→ Better labour use – Crews focus on pruning and picking instead of inspecting traps.

→ Clearer trends – Continuous data highlights hotspots early and helps break the pest cycle.

Healthier trees, fewer unnecessary sprays, and sharper resource use — all season long.

What is your integrated pest-management strategy, and how do you monitor the presence of SWD?

Petiole

15/03/2026

29/12/2025

11/12/2025

10/12/2025

09/12/2025

08/12/2025

07/12/2025

06/12/2025

23/11/2025

12/10/2025

16/07/2025

Address

Website

Alerts

Contact The Business

Shortcuts

Share