Search the GenAI Directory

Odd readings in crop data often surface too late, after someone finally opens the spreadsheet. This is an importable workflow file for a no-code automation tool that flags unusual values in crop datasets as they come in.

Spotting and naming crops and pests in greenhouse images needs varied labelled examples. This is a published crop vision dataset for classification and detection, described in a peer reviewed article.

Predicting how a plant will perform from its genetic markers usually gives you a number with no explanation. This is code and a web tool for crop trait prediction that also points to which markers drove the result.

Every new field, season or camera seems to need fresh labelled images, and labelling is the bottleneck. This is code for a model that generates labelled crop imagery from existing pictures so you can train detectors without annotating again.

Sometimes you want a light crop and soil water calculation to describe growing conditions, not a full modelling suite. This is Python source code with simple growth and water balance routines you can read and adapt.

Before trusting an AI system on a farming or food task you need a way to check it. Crop and crop type classification from satellite, climate and elevation inputs, with region specific evaluation tasks and baseline implementations.

Recording a field visit and then writing up recommendations doubles the work. This is an agent inside a farm management app that takes in photos or voice notes, carries out tasks with its own tools and drafts reports by conversation.

Sourcing, agronomy and risk questions usually mean pulling numbers out of several prediction systems by hand. This is a hosted agent layer for food and farming that other software can reach through a plug-in interface for AI assistants.

Sourcing and production teams often need several analysts to combine weather, soil, satellite and crop-model data before they can compare regions or future scenarios. Cropin SAGE is a hosted GenAI query and visualization platform for asking global crop-production, suitability and climate-risk questions in plain language.

Getting a language model to handle crop science needs both training examples and a matching test. This is a downloadable pair of resources, an instruction dataset on crop science and a test set built alongside it, with code.

Linking weather, satellite views and harvest outcomes is normally a lot of plumbing before any modelling starts. This is a joined dataset of climate records, satellite imagery and crop yields, with Python code for downloading and loading it.

Trial data from different crops and teams rarely lines up, so merging it turns into guesswork. This is a shared vocabulary and lookup service for crop traits, measurement methods and scales, available as a website and for software to query.

Understanding how a crop performed across a field usually means pulling imagery from several sources. This is a published dataset of aerial and satellite views tied to crop performance, downloadable for model training or testing.

Deciding what to plant where, and how much to spend on it, is a yearly puzzle with a lot riding on it. This is a listing in a cloud provider’s agent directory for a hosted crop planning tool aimed at growers and advisers.

Season planning and harvest timing raise the same questions every year, written out again from scratch. This is a packaged set of prompts for crop planning and harvest work with a chat assistant.

You want an AI assistant to talk about European crop yields, prices and risk without you fetching the numbers first. This is a plug-in interface for AI assistants that pulls together soil, climate and market figures and returns forecasts.

Deciding when to fertilise or irrigate is a sequence of choices, and testing strategies in the field takes years. This is a training environment where software agents learn management decisions by practising against a crop growth simulation.

Sometimes an agronomy question needs a back and forth conversation, not a search result. This is a preconfigured chat assistant, opened through a shared link, for talking through crop science and agronomy topics.

Farm advice tools sometimes have to work without a reliable internet link. These are downloadable model weights, with code, for a language model geared to agriculture that you can run on your own machine.

Farm records exist, but turning them into decisions about the next crop is the hard part. This is a preconfigured chat assistant, opened through a shared link, that reviews farm data and suggests changes to yields.

You have field photos and want plain answers about what is in them. This is a downloadable model that reads pictures alongside written questions about crops, shipped with code so you can run it yourself.

Advisors already keep field records in an agronomy platform but still have to interpret them themselves. This is a chat assistant layered on that platform so questions can be asked in ordinary language.

Growers using a farm platform often want a quick answer instead of clicking through screens. This is a chat assistant built into a Brazilian grower application, answering questions inside the tool they already log into.

Your app needs agronomy answers such as disease risk or growth stage without you building the science behind them. These are developer interfaces returning relative yield, stress, disease risk, growth stage and seeding rate suggestions.

Farmers and extension staff often need a crop answer that is actually backed by university guidance, not a forum post. This is a free web assistant that searches university farming material, handles photos as well as text, and does some calculations for you.

You are building a grower facing product and do not want to gather soil and field readings yourself. This is a web connection point that feeds field, crop and climate information plus agronomic notes into someone else’s software.

You want ideas for getting more out of a field while keeping the ground in good shape. This is a preconfigured chat assistant, reached by a shared link, that suggests crop and sustainable practice changes.

Running a subscription vegetable box scheme means planning plantings, shares and member communication all together. This is a packaged set of prompts for community supported agriculture planning and day to day operations.

Your agrifood company has data in many places and no team free to build agents on top of it. This is a consulting and engineering house that builds custom AI agents and runs its own data platform for clients.

Before trusting an AI system on a farming or food task you need a way to check it. Predicts marine fish species for held-out test images, built on the FishNet dataset.

Before trusting an AI system on a farming or food task you need a way to check it. Tracks sweet pepper fruit across RGB-D robot sequences with heavy occlusion and illumination variation, scored using HOTA.

Counting leaves and outlining them by hand across hundreds of plant photos is tedious. This is a downloadable set of top down plant images with each leaf traced, first used for a research contest.

Yield forecasts are hard to compare when everyone uses their own data and test setup. This is an open benchmark for maize and wheat yield forecasting across many countries, with Python code you can run yourself.

Rotations change soil water and nutrients over many years, which is hard to reason about from one season of records. This is a simulation model covering crop rotations along with water, carbon and nutrient movement in the soil.

Balancing a dairy ration around lactation stage, what is in the shed and what feed costs takes time you may not have. This is a copyable prompt that guides a chat assistant through putting that mix together.

You want a daily ration for a cow of a given weight and milk target, using feeds available nearby in Ethiopia. This is a prompt you paste into a chat assistant to get that ration written out.

Picking up signs of trouble from the sounds cows make is hard to do around the clock. This is an openly shared dataset of sound features taken from dairy cow calls, for people building audio recognition tools.

A lender or partner wants numbers on a dairy investment and building the spreadsheet takes days. This is a copyable prompt that guides a chat assistant through a dairy farm model of costs, income, stock and profit.

Dairy herd data shows up in several systems, so patterns get noticed late or not at all. This is a chat assistant that joins that herd data together, points out likely problems, explains what it sees and suggests what to do next.

Smallholders who cannot easily read farming advice, or who have an unfamiliar plant problem in front of them, need another way in. This is a farming assistant you can listen to, talk with or show a plant photo to, in many languages.

Working out when dairy cows eat and chew usually needs someone observing or a costly sensor. This is an open audio collection of many hours of cow sounds recorded in pasture and in barns.

If you run a farm credit assistant in Portuguese, you need something to catch off-topic and manipulative prompts. These are downloadable weights for a text classifier, paired with a second model as a checker, trained on made-up attack examples.

Quotes, invoices, rosters and customer messages in a flower business all need attention at once. This is a hosted assistant over that shop’s records that drafts text, summarises, produces reports and runs multi-step tasks.

Building anything that understands cattle farming needs matched records rather than loose files. This is a downloadable dataset about cattle keeping that mixes text and images, published for training and grounding models.

Before trusting an AI system on a farming or food task you need a way to check it. Object detectors on multi-class weed detection in cotton production, comparing YOLO family models on a shared annotated image set.

Before trusting an AI system on a farming or food task you need a way to check it. Image classification of fertiliser treatments to recognise nutrient deficiencies in winter wheat and winter rye plots.

Soil and vegetation maps mean little to a grower who cannot read the colour scales. This is a hosted web tool that looks at those maps and writes a plain description of what they suggest for the field.

Roots are hard to see, so root studies are short on data to work with. This is a code library that generates synthetic root systems and fits them to observations, giving you root architectures to experiment with.

Drafting farm reports or step by step explanations with a general model often needs extra prompting. These are downloadable model weights, with code, for a reasoning focused language model further trained on agricultural material.

If expert farm advice is hard to come by, a general assistant is often the first thing people try. This is an openly available general purpose language model, callable through a programming interface, that has been run against agricultural advisory questions.

Sometimes you need a general text model to draft or reason over food chemistry material rather than a specialist one. This is an openly released language model, usable through a website, that has been tried on food chemistry test questions.

Identifying weed species from a field photo is a step towards targeted spraying. This is a downloadable collection of Australian weed images with species labels, alongside code and trained models for running it.

Working out where one field ends and the next begins is tedious to draw by hand. These are downloadable model weights that outline individual field parcels in satellite pictures, whether the imagery is coarse or fine grained.

Writing up soil findings and explanations takes time that agronomy staff rarely have. These are downloadable model weights, with code, for a text generating model geared towards soil topics.

Reviewing farmer advice answers by hand does not scale, and you still need to catch contradictions and vague replies. This is a set of ready-made prompts for checking advisory answers on points like factual accuracy, specificity and relevance.

Teams building farm-facing tools end up rebuilding the same agricultural know-how and task flows each time. This is a hosted platform that offers farming specific assistants and ready-made workflows you can put to work or build on.

You want a model that can look at a crop photo and describe what it sees in farm terms. These are downloadable adapter weights, with code, that adapt an existing text and image model to crop and field pictures.

Voice services for farmers only work if the software hears farming words correctly. This is a test dataset of agricultural speech recordings with transcripts, for measuring how speech to text systems handle the vocabulary.

Photos of sick plants are only useful for training once someone knowledgeable has labelled them. This is a downloadable set of crop disease images with labels added by people with subject knowledge.

Useful facts sit inside long advisory documents and someone has to pull them out. These are downloadable adapter weights, with code, for a small language model that lifts individual facts out of text.

Once a system pulls out facts, you still need to know which ones actually matter for the question asked. These are downloadable adapter weights for a small language model that rates how relevant a fact is.

Knowing what growers actually ask is the starting point for designing an advice service. This is a large downloadable set of farming questions and answers gathered from a chat service for farmers.

Automatically scraped answers are shaky ground for teaching an advice assistant. This is a downloadable set of farming questions and answers that people went through and tidied by hand, with code alongside.

Automatic scores do not always match what a person would prefer to read. This is a dataset of human judgements comparing pairs of answers, so you can tune or check automatic scoring against real preferences.

An assistant handling messages needs to work out what the person is asking for before it can respond. These are downloadable adapter weights, with code, for a small language model that sorts incoming requests into intent categories.

To serve Kenyan growers well you first need to see the questions they really send. This is a downloadable set of anonymised farmer queries collected from an early version of a messaging advice bot.

When an assistant looks things up in documents before replying, how the text is cut into pieces matters. This is a downloadable dataset used to examine those text chunks, published with code for people building such systems.

Working out which insect is in a field photo usually needs a specialist or a lot of guidebook flipping. These are downloadable model weights, with code, that sort pictures of farm insects into named categories.

Sorting leaf photos by disease is a job people still do by eye. These are downloadable image model weights tuned to classify plant diseases from pictures, usable on their own or as a base for scoring disease severity.

Starting a plant image task from a general purpose vision model means it has never really looked at plants. These are downloadable image model weights already tuned on a large plant identification photo collection, meant as a base for plant vision work.

Plant measurement research often needs code you can adapt rather than a finished tool. This is a code release related to plant measurement work, published on a public repository for others to read and build on.

Drone photos of fields pile up and nobody gets round to turning them into answers. This is an automation workflow blueprint that stores and indexes the imagery, then lets you ask about crop condition in plain language.

Comparing management or climate scenarios across many crops needs a model of what happens in the field. This is a process based cropping system model covering more than forty crops and the links between soil, plant and air.

Getting an older crop model to build and run the same way on every machine eats hours. This is a container recipe that packages the model with its dependencies so batch and experiment files run the same everywhere.

A public proof-of-concept MCP server that wraps the DSSAT crop-simulation model so an LLM agent can list model inputs, run crop simulations, conduct sensitivity analysis and evaluate calibration statistics.

Farm and trading software often needs weather, price and crop information from several vendors at once. This is a commercial catalogue of programming interfaces covering weather, markets and agronomy that developers can subscribe to and call from their own systems.

Spotting a crop problem from the ground can happen too late, and satellite data is fiddly to assemble. This is a data interface offering field imagery, weather and crop condition measures for your own monitoring tools.

Checking food packaging against Japanese labelling and advertising rules is slow line by line work. This is a hosted service where several cooperating agents review chosen checkpoints and report whether the wording meets those legal requirements.

Growers moving to organic methods have plenty of questions and few advisers who specialise in them. This is a web chat assistant that looks up organic farming material before answering and offers suggestions for your situation.

Judging whether on-farm sensor systems are worth it means having measurements to compare. This is a dataset of performance indicators from sensor based livestock monitoring, published in an open catalogue for download.

Questions about what might be unsafe in a food product usually mean digging through reports by hand. This is example code for a question answering app that searches food hazard documents before it replies, so developers can build on it.

Questions about crop protection products and how to use them often need an agronomist on hand. This is a small language model, tuned for crop protection questions, that can be deployed from a cloud model catalogue with code provided.

A 308M-parameter multilingual text embedding model for retrieval, similarity search, clustering and classification on everyday devices. Google documents quantized operation below 200MB of RAM and offline generation of embeddings. It is a general retrieval component rather than an agrifood-specific model.

Brand and category teams wait on analysts to pull retail sales numbers and turn them into slides. This is an agent you reach through a programming interface or a plug-in interface for AI assistants, which queries retail measurement data and drafts the write-up. The publisher describes it in the context of Circana Liquid AI.

Leaf image work needs example pictures, and gathering them yourself is slow. This is a hosted dataset of leaf and plant photographs you can preview online and download for image recognition tasks.

Recognising which insect is chewing a crop takes a stock of labelled pictures. This is a downloadable dataset of insect pest records, meant for training, grounding or testing farm related models.

Carbon reporting for livestock needs emissions figures you can trace to a source. This is a downloadable table of methane from animal digestion across African countries, taken from official statistics and packaged for code.

Building a farm app means finding field boundaries, crop condition and weather history from somewhere. This is a data service that hands your software satellite readings on crop growth, field outlines, forecasts and past weather records.

Keeping sheep records tidy and running flock, mating and inbreeding checks by hand is slow work. This is a copyable code repository showing a farm workflow that connects a coding assistant to a sheep breeding records service.

Language models can give shaky answers on farm exam style questions, and you want more confidence. This is a prompting method that asks a model several times and refines its answers, described in a study and usable online.

Deciding when to spray or irrigate means checking forecasts, soil moisture and water loss figures in several places. This is a plug-in interface for AI assistants that fetches weather, forecasts, soil conditions and field alerts from a public weather service.

Proving that a farm plot is not linked to deforestation means reading a European regulation and checking map data. This is a public assistant that searches the regulation before answering and can also analyse plot boundary files you upload.

Working out what the European deforestation rules ask of you means wading through the legal text. This is a custom chat assistant set up on that regulation and related documents, so you can ask about it in plain questions.

Proving that your supply is free of deforestation means reading supplier paperwork and land records line by line. This is a hosted compliance tool that puts AI agents and locally run language models over those documents.

Dairy and livestock data lands in several systems and turning it into an actual change on farm takes chasing. This is a hosted tool that watches the operation, explains what it suggests and carries out the changes you approve.

Groups working on staple crop genomes want a general DNA model they can adapt rather than build. This is a downloadable generative genome model covering life broadly, offered in three sizes, which one crop team has already tuned for staple crops.

Running a very large genome model yourself means wrangling hardware before you get any sequence out. This is a packaged service that serves that model over a web interface, on your own machines or hosted, and can load your own tuned versions. The publisher describes it in the context of NVIDIA BioNeMo microservice + hosted API.

You want to see what a genome model produces before wiring it into your own code. This is a free web page where you type in a starting sequence and a species and get generated DNA back.

PyTorch’s open on-device inference runtime for exporting and running machine-learning and language models through C++, Swift and Java bindings. It supports mobile, desktop and embedded targets with hardware-specific backends from multiple chip vendors.

Growing tropical or unusual fruit means guessing at soil, climate and care because local guidance rarely covers it. This is a preconfigured chat assistant, opened through a shared link, focused on rare and tropical fruit growing.