Search the GenAI Directory

Measuring plant traits from images means stitching together several tools and remembering how each is run. This is code for an assistant that takes a plain request, picks the tools, runs the analysis and returns the measurements.

Comparing crop and weed vision methods fairly needs one shared set with careful labels. This is a downloadable field image benchmark marking crops, weeds, individual plants and leaves, plus extra unlabelled images for pretraining.

Sometimes you want watering answers from a model small enough to run on modest hardware. This is a downloadable compact language model tuned to answer questions about farm irrigation.

A 3.8B-parameter open small language model for instruction following, reasoning and function calling in compute-constrained environments. Microsoft documents on-device use and ONNX Runtime optimization. It is general-purpose and needs agrifood-specific grounding and evaluation.

A 5.6B-parameter open multimodal model that processes text, images and speech in one architecture. Microsoft describes it as optimized for low-latency on-device and edge execution. It is general-purpose and not trained or validated specifically for agrifood.

Pig producers ask health, feed and housing questions that general assistants handle poorly. This is a hosted assistant built around swine production, also used inside a wearable headset for hands-free help in the barn.

Following pig numbers in African countries over the years means pulling from many yearly releases. This is a downloadable table of pig head counts by country and year, formatted for direct use in code.

Grazing and pasture decisions depend on your own paddock history, the weather and what the grass is actually doing. This is an assistant inside a farm app that pulls together records, satellite pictures and weather to answer grazing and feeding questions.

Watching fields change over a season needs pictures from above on a regular schedule. This is a set of data interfaces for satellite imagery and derived measures, with a farming trial and commercial plans available.

Smallholders deciding when to water often have no field sensors to go on. This is a messaging based advisory service that sends watering and weather guidance twice a week along with alerts for crop stress and severe weather.

Working out which parts of a crop genome matter is slow going one gene at a time. This is a set of downloadable plant genome models, offered in three sizes, trained across sixty five flowering plant genomes.

Breeding programmes need some way to guess which plant genetic changes matter before running field trials. These are downloadable model weights and code for a plant DNA model that produces sequence representations, scores variants and can be tuned for prediction tasks.

Recreating or extending a plant genome model means first gathering genomes and labels from many sources. This is a packaged download of sixteen flowering plant genomes plus a cross species set of single letter annotations used to train and check such a model.

Finding individual sugar beet plants in row imagery is a common step for weeding machines and plant measurement research, and it needs marked-up examples. This is a set of colour and near-infrared field images of sugar beet rows with plant positions labelled.

Before trusting an AI system on a farming or food task you need a way to check it. Multi-label identification of multiple plant species in images of vegetation plots.

You want to try training your own model to spot plant diseases in leaf photos but need somewhere to start. This is a code project with sample image data and training scripts for classifying plant disease from pictures.

Predicting how a stretch of plant DNA behaves normally means building a model from scratch for each question. This is a family of downloadable models trained on plant genomes, each with ready made versions for tasks like spotting promoters and gene switches.

Lab photos of diseased leaves do not look much like what you see in a real field. This is a downloadable set of plant disease images taken in the wild across a range of species and conditions.

Drawing boxes around diseased plant parts by hand is slow work to start from. This is an openly listed download of plant disease images with object boxes marked, for training or scoring detection models.

Answering specialist plant science questions from a photograph is a high bar for any assistant. This is a downloadable test set of plant images paired with expert level questions and answers for scoring models.

Someone releases a plant genetics model and you want to know what it can actually do. This is a downloadable set of test tasks for plant genome models, published as the evaluation companion to one such model.

Different model designs suit different genome questions, and the options for plants are limited. This is a downloadable plant genome model using a hybrid design, trained on sequence from many plant species.

Plant gene lookups mean visiting half a dozen separate databases and learning the quirks of each. This is a plug-in interface for AI assistants that gives them around fifty tools for querying plant gene, protein, trait and paper collections in one place.

Studying long stretches of plant DNA at once is hard when a model only reads short windows. This is a downloadable plant genome model that takes in longer sequences, released with ready made versions for nine different sequence tasks.

Training or checking a plant genome model needs labelled sequences, which are tedious to assemble yourself. These are downloadable training and test sets covering plant gene control regions and genome labels, prepared to go with a plant DNA model.

Diagnosing a plant problem is rarely one question, it is a back and forth that narrows things down. This is a large test set of plant images with multi-step diagnostic questions and answers for checking models that read pictures.

If you are training your own plant DNA model, collecting and cleaning genomes from many species eats weeks. This is that groundwork already done, a downloadable corpus of plant genomes packaged for model builders.

To use an AI designed gene editor in rice you need it packaged in a form plant cells will read. This is a set of orderable plant expression constructs carrying that editor, including versions that make single letter DNA changes.

Telling apple leaf diseases apart from a picture is hard without labelled examples to learn from. This is a downloadable image set of apple leaves sorted by condition, originally released for a public contest.

Before trusting an AI system on a farming or food task you need a way to check it. Identifies the category of foliar diseases in apple trees from leaf images.

Pre training a plant DNA model starts with a big pile of genome sequence, which is a chore to collect. This is the downloadable reference genome collection that was used to pre train a family of plant DNA models.

Plant RNA questions often start from scratch because there is no shared model to build from. This is a small downloadable model trained on RNA from over a thousand plant species that you can adapt to your own RNA tasks.

Before trusting a model that outlines plants in photos, you need a fixed set to test it against. This is a downloadable collection of about 1,200 plant images with outlines, packaged for a common review tool.

Anyone starting on crop disease recognition needs labelled leaf photos, and gathering them yourself takes a season. This is an open image collection of healthy and diseased leaves across fourteen crops and their common diseases.

Before letting an app diagnose crop disease from a photo, you want a way to test its answers. This is a question and answer dataset built on leaf images, for scoring assistants that work with text and pictures.

Food safety records have to be written, filed and kept current, and it usually falls to whoever has a spare hour. This is a hosted workflow inside a food traceability platform that creates and looks after those compliance records.

Pesticide labels are full of short formulation codes that a model will simply misread without a glossary. This is a small downloadable reference list of pesticide formulation types with definitions in Chinese and English.

Answering questions about which crop chemicals are approved and how means reading registration paperwork line by line. This is a downloadable set of Chinese pesticide registration records with fields like product, active ingredient, formulation, toxicity, holder and validity dates.

Plant science answers are scattered across a century of journals that no one person can read. These are downloadable language models further trained on more than a million plant science papers, with versions set up to follow instructions.

Breeding teams often have years of data and no data specialist free to analyse it. This is a hosted platform where an AI agent frames the analysis, runs the genetics tools on a cluster and talks through what came out.

Potato work spreads across thousands of papers and a pile of separate sequence tools. This is a hosted question answering site that searches potato research before answering and bundles in sequence search, gene labels and a genome viewer.

Dispatch, invoicing and stock counts eat up hours on a commercial poultry operation. This is a build team you hire to set up AI assistants and automated workflows that take over those repeating office jobs.

Poultry numbers across African countries are scattered across yearly statistical releases. This is a downloadable table of poultry head counts by country and year, arranged so a model or script can read it.

You need a long run of poultry numbers for Asian countries without assembling it by hand. This is a downloadable table of bird counts by country and year, packaged for direct loading into analysis.

Comparing poultry flock sizes between European countries over time takes figures in a single tidy file. This is a downloadable table of poultry counts by country and year, ready to load into code.

Getting useful farm answers out of a chat assistant depends a lot on the wording you use. This is a published pack of ten prompts for precision farming and modern farm management, written for common chat assistants.

You are considering guidance systems, variable rate work or field mapping and want to sound out the options. This is a preconfigured chat assistant, shared by link and offered alongside support, focused on precision farming questions.

You have an agricultural vision model and no consistent way to check how it behaves across different tasks. This is an installable Python toolkit that runs image search, region labelling and diagnostic checks so results can be compared.

Setting up a test run means downloading images, sorting them into groups and building comparison pairs by hand. This is a companion Python package that fetches the drone stress test images and assembles the comparison sets for you.

Agricultural image archives are hard to search by meaning. Precision AI’s hosted, OpenAI-compatible service turns one image or a batch into embedding vectors and dense patch tokens, with an Explorer for inspecting results and AgriViz for browsing agricultural image collections.

You keep hearing about generative AI and want a guided start rather than trial and error. It is a short remote course over three sessions covering prompting, chat assistants, note taking tools and the risks, with farming examples.

Grain and oilseed decisions rest on government reports and positioning data that take time to gather. This is a hosted plug-in interface that brings official crop reports, export sales, trader positioning and sector spreads to an AI assistant.

Adapting satellite imagery to a new crop mapping task usually needs more labelled examples than you have. These are downloadable model weights, pretrained on large amounts of Earth observation imagery over time, ready for you to tune.

Building crop or land use maps from satellite pictures usually means starting from scratch. This is a pretrained set of model weights for satellite imagery over time, plus code to fine tune it and run it.

This Hugging Face organisation hub is a catalogue of IBM-NASA geospatial models, datasets and demos, including Prithvi-family resources. It is a discovery page rather than a single downloadable asset, and each linked item has its own access terms and licence.

Government procurement teams lose time searching separate contract databases and reading long solicitation files before they can draft a new package. ProcureSight is a hosted GenAI procurement workspace that searches public acquisition data, answers questions with source links and helps create requirements documents.

Sometimes you want a chat assistant that answers like an experienced agronomist rather than a generalist. This is a paid marketplace prompt that sets up that persona inside a chat assistant you already use.

Designing a new enzyme for crop traits or farm inputs from first principles is hard going. These are downloadable generative protein models in six sizes that propose new protein sequences and can be tuned on your own protein families.

Choosing a model for farm image tasks is easier when someone has already written the exam. This is a downloadable evaluation dataset of agricultural questions and images, published for comparing model answers.

Sorting coffee beans by grade takes trained eyes, and there is little labelled material for teaching a computer to do it. This is an image collection of coffee beans with quality labels, ready for training or testing vision models.

Spotting disease on cotton leaves early is hard when you have no labelled pictures to train on. This is a set of cotton leaf photographs sorted by disease type, meant for building or checking image recognition work.

Telling weeds apart from cotton plants in field photos is fiddly work for a person and needs examples for a machine. This is a labelled picture collection of cotton fields with weeds marked, for training detection models.

Researchers building plant recognition tools often run short of labelled images across many species. This is a very large photo collection drawn from a nature observation project, covering thousands of crop and weed species with species labels.

Counting maize tassels by hand across a field plot is slow and easy to get wrong. This is an image set of maize plants with tassels marked up, so you can train software to find and count them.

Growers and researchers want to catch tomato problems before they spread, but training material is scarce. This is a labelled collection of tomato plant images grouped by disease, useful for building or testing recognition models.

Deciding where to prune vines is skilled work, and there is not much labelled imagery for teaching software to help. This is a picture set of grapevines with pruning points and structure marked for training vision models.

Estimating soybean yield means counting pods, which is tedious and hard to do at scale. This is an image collection of soybean plants with individual pods labelled, so detection models have something to learn from.

Protein research involves both making sense of existing sequences and designing new ones. These are downloadable model weights and code for a protein sequence model that can describe proteins and generate new sequences to a specification.

Teams working on proteins often want to train or adapt a sequence model rather than start from nothing. These are downloadable model weights together with code for pretraining and further tuning on your own protein data.

Working out which crops suit which land across a region is a long manual exercise with maps and climate tables. This is a Python code library that carries out land suitability and productivity zoning calculations for you.

Running a water and yield model by editing text files and clicking through a program does not scale. This is a Python wrapper that writes the settings and runs the underlying crop water model from your own code.

You would rather set up and launch crop model runs from a notebook than from legacy input screens. This is a Python interface that prepares the model’s input files and runs simulations for you.

Shading from solar panels over a field changes how the crop underneath grows, and that is hard to predict. This is a slimmed down Python version of an established crop model, set up for panel shading questions too.

Q is an AI assistant inside QuickTrials that answers natural-language questions about field trials, finds gaps or unexpected measurements, ranks subjects and suggests improvements to experimental design.

Before trusting an AI system on a farming or food task you need a way to check it. Certifies agricultural index insurance products by measuring basis risk between index estimated losses and actual farmer losses.

Crop protection, seed and fertilizer research slows down when data and analysis steps live in separate places. This is a paid implementation service that sets up a no-code agricultural research environment linking those datasets and scientific pipelines, with a language model connected.

After a commodity trade is agreed, shipping documents have to be matched against instructions and bank paperwork line by line. This is a hosted tool that sorts those documents, checks them against the instructions and points out where they disagree.

Not every farm advice project can pay for a hosted model, so openly available ones get a look. These are downloadable model weights, also callable through programming interfaces, for a general purpose language model checked on agricultural advisory questions.

The small 2504 release group from Alibaba’s open-weight Qwen3 language-model family, covering 0.6B, 1.7B and 4B variants. These models support reasoning and non-reasoning modes, tool use and more than 100 languages, with documented local and mobile runtime paths. They are not agrifood-tuned and do not represent every current Qwen3 release.

You need a farming question and answer setup that draws on your own documents instead of guessing. This is an open code pipeline that retrieves matching farm reference passages first and then has a language model answer from them.

Your farm and research data is messy, inconsistent and impossible to cite when an assistant answers from it. This is a delivery engagement of a few months that tidies and checks that data, then grounds answers in it with sources shown.

Customers and staff ask questions your documents already answer, but nobody can find the right page. This is a service that builds document searching assistants and internal knowledge bases shaped for farming and food technology companies.

Brazilian producers already message on their phones, so that is where a technical question is most likely to be asked. This is an assistant delivered through a messaging app that handles routine agricultural guidance requests for extension services.

Preparing a farm or company for a sustainability certification audit means knowing what evidence assessors want. This is a preconfigured chat assistant, opened through a shared link, that works through the Rainforest Alliance farm standards.

Crop chemical work means keeping up with papers, your own protocols and shifting registration rules all at once. This is a hosted chat assistant for farm chemistry that reads across those sources, cites the line it used and watches for regulatory updates.

Getting a prototype off the ground is easier when the text model already knows farming language. These are downloadable model weights, with code, for a text model trained on agricultural writing that you can build features on.

Regulatory teams drown in updates in several languages and still have to answer the business quickly. This is a hosted assistant that answers regulatory questions with citations, and behind it, tooling that translates, sorts and summarises incoming regulatory material.

You want to build soil carbon and organic matter, but the practices and claims are hard to sort out. This is a preconfigured chat assistant, opened by link, discussing regenerative farming, soil health and carbon storage.

Shifting how you farm towards regenerative practice raises plenty of practical, on the ground questions. This is a preconfigured chat assistant, opened through a shared link, written as a grower’s guide to regenerative agriculture.

Moving to practices that build soil life and variety takes years and steady guidance along the way. This is a preconfigured chat assistant, opened through a shared link, supporting growers who are adopting regenerative farming methods.

Teams working on regenerative farming and carbon reporting often cannot find enough records to train or test on. This is a set of part-generated farm records, packaged so you can train a model on them or use them as reference material.

Taking one product into several markets at once means repeating the same label and ingredient checks for each set of rules. This is a hosted tool that runs those checks together and returns a per-market verdict with the gaps it found.

Launching a food or supplement in a new market raises rule questions nobody in-house can answer quickly. This is a hosted advisory tool that answers those questions across many countries and cites the regulations it drew on.

Before trusting an AI system on a farming or food task you need a way to check it. Requires soil carbon models to meet bias limits and 90 percent confidence coverage of 90 percent prediction intervals, checked by expert verifiers.

Anyone building coffee tools needs field data they are allowed to reuse. This is a coffee dataset credited to Producers Direct and Alliance Bioversity and CIAT, downloadable under a CC BY-SA 4.0 licence.

Teaching software to outline rice plants and their parts in photos needs pictures where someone has already drawn the outlines. This is a downloadable rice field image dataset with those outlines marked, so you can train models to do the same.

Buying and moving a crop involves grading, haggling, paperwork and trucks, mostly handled by phone and spreadsheet. This is a hosted platform for farmers, buyers and logistics providers that assesses crop quality and keeps negotiation, sales and transport steps in one place.

Tracing a food item from farm gate through wholesale to shelf price means visiting several separate public sources. This is a connector giving AI assistants a set of food price lookup tools plus recall records, chained across those stages.

Finding disease resistance genes in a newly sequenced crop genome is fiddly, manual annotation work. This is a code pipeline built on protein language models that scans a plant genome and labels resistance genes across fifteen structural types.

Before trusting an AI system on a farming or food task you need a way to check it. 2,935 Russian multiple-choice questions on botany, crop production and genetics testing agronomy knowledge acquired in pretraining.