GREEN24AI · Development

From photo to reliable care advice

GREEN24AI checks your plant with several independent modules – and tells you honestly how certain the result is. Here you can see how the system has grown and where it is heading.

Status: version V139 ·

How GREEN24AI thinks

  • Pre-check – the green firewall.

    Detects whether a plant is really in the picture, checks the photo quality and gives a first indication of the plant type. For blurred photos you get tips instead of an uncertain result.

  • The analysis modules: BotaniA, BioloB, PlantiC and VerdE.

    They assess the same photo independently of one another – plant, damage pattern, care. A single opinion is not enough.

  • FloriD – the identification specialist.

    Joins in when the modules do not agree on the plant.

  • Step-by-step verification.

    What counts is what two independent sources agree on. If that is not enough, GREEN24AI checks again in a targeted way – or asks for another photo and says what it should show.

  • Safe and honest.

    Fixed rules from GREEN24’s gardening practice: safety notes right at the work step, check first and then cut, a pathogen is only named once it is confirmed.

  • Learning.

    Reviewed cases flow back: as guidance for our modules and into our GREEN24AI, which keeps learning in the background.

What is new

  1. 0.9.38

    GREEN24AI shows its development

    From now on, this page always shows the status of the running version – with everything that is new.

  2. 0.9.37

    Result at a glance

    First the steps for today, then when you need help, then matching products – with a note on why they are shown.

  3. 0.9.36

    Fine-tuning

    Neutral findings are named more clearly. Cases in which the modules disagreed are reviewed by G24AI first.

  4. 0.9.35

    Step-by-step identification

    What two independent sources confirm is what counts. Otherwise GREEN24AI checks again or asks for one more specific photo.

  5. 0.9.34

    Safety right at the step

    Safety notes appear where you act. And the rule is: check first, then cut.

  6. 0.9.33

    Safe and honest

    A pathogen is only named when it is confirmed. GREEN24AI only recommends feeding at the right time.

  7. 0.9.32

    Traceable

    Every result carries the version it was created with.

  8. 0.9.30

    Better photos, better results

    For blurry, dark or distant photos you get tips for a better picture instead of an uncertain result.

  9. 0.9.26

    Toxicity in levels

    Assessed separately for people, pets and wildlife.

  10. 0.9.21

    Smart follow-up questions

    If the modules disagree, GREEN24AI asks specific follow-up questions.

  11. 0.9.13

    Greenkeeper

    A dedicated module checks lawns – for weeds, nutrient deficiency, diseases and care.

  12. 0.9.12

    Pre-check

    The Green Firewall checks every photo first: does it show a real plant, and is the image quality good enough?

The journey so far

  1. The beginning: the plant consultation.

    One photo, one assessment.

  2. Majority instead of a single opinion.

    Several modules assess the same photo; “Recently checked” shows anonymously what GREEN24AI recognises; every result gets a link for sharing.

  3. Knowledge grows along.

    Knowledge base with consent, review list, learning list against known mix-ups; season and region flow into the analysis without being stored.

  4. Pro analysis.

    In-depth second analysis with health score, chain of causes, care plan and plant profile.

  5. Greenkeeper and pre-check.

    A dedicated module for lawns; the pre-check sorts out non-plant images and watches the photo quality.

  6. Follow-up questions and feedback.

    GREEN24AI asks targeted follow-up questions when the modules disagree; users rate the analysis; for every analysis you can ask our plant experts.

  7. Safe, honest, clear.

    Safety notes at the step, step-by-step identification, toxicity in levels, a result page with the most important things first and products that say why they are shown.

For experts

  • Ensemble of independent modules with a weighted majority decision and a grace period; a specialised identification module as an additional source.
  • Separate confidence for plant (species, genus) and finding; verification details only from the actual votes.
  • Rule layer after merging: catalogue of 30 damage patterns, verified action steps and fixed safety notes – no freely worded safety sentences.
  • Every version passes automatic tests (currently more than 2,700) and comparison runs on sample cases.
  • Privacy by design: photos without metadata, storage only with consent; the GREEN24AI service and G24AI, our own AI, run on our own servers in Germany.
  • Every result carries the version of service and interface – the basis for traceable learning.

Non-binding

Outlook

What we are working on:

  • Woody plant pruning module
  • Cropping tool for the damaged area
  • G24AI gradually takes on more tasks
  • Interfaces for partners (app, plugin)

Order and dates may change.

Master – GREEN24AI for your project

We build Master individually: as an app or plugin for your website, for garden centres and nurseries, for research and professional use. You tell us what your users should know – we align modules, rules and interface with it.

Request Master

This takes you to the GREEN24 contact page, topic “GREEN24AI Master”. Please briefly describe your project there.

For partners and investors: get in touch with us.

Note

This page describes the state of development and non-binding plans. It is not a commitment to features or dates and not an offer to acquire shares. GREEN24AI provides assessments and does not replace professional advice.