Reference
Glossary
Why this page exists
These are the definitions used on every other page of this site. Words like detection, threshold, and abstain carry different meanings across agronomy, machine learning, and equipment engineering; fixing one meaning each is what lets the rest of the site say something precise.
- Published
- Last updated
- Terms
- 37
Computer vision
- Adjudication
- Review of a disputed or uncertain biological label by a qualified reviewer, producing a recorded resolution rather than an averaged guess. Adjudication decisions are retained as part of dataset lineage.
- In depth:Field data verificationDecision traceability
- Agricultural computer vision
- Also: machine vision, agricultural machine vision
- The application of computer-vision models to imagery captured in working crop fields in order to identify plants, pests, life stages, symptoms, and their location within the frame. It estimates what is visible in an image; it does not by itself establish whether an action is justified.
- In depth:Agricultural computer vision
- Annotation
- A human-applied label on a field image — species, life stage, boundary, or quality judgment — used to train and evaluate models. Annotations that involve biological judgment are adjudicated rather than accepted from a single labeler.
- In depth:Field data verification
- Calibration
- Two related things: the camera calibration that maps image pixels to field geometry, and the confidence calibration that determines whether a model's stated confidence corresponds to its observed reliability. Both are recorded per decision.
- In depth:Field data verificationUncertainty and abstention
- Detection
- A model output asserting that a target — a pest, a life stage, a symptom — is present at a location in an image, usually with a confidence value. A detection is an estimate about an image, not a statement about a field.
- In depth:Agricultural computer visionBiological threshold policies
- Edge inference
- Running a model on hardware in the field — on a vehicle, implement, or handheld device — rather than sending imagery to a server first. Edge inference constrains model size and compute budget and changes what evidence can be retained.
- In depth:OEM integration
- Model version
- The specific, immutable identifier of the trained model that produced a detection, together with the dataset and training run behind it. Without a recorded model version, a past result cannot be reproduced or re-examined.
- In depth:Decision traceability
Decision layer
- Abstention
- Also: abstain
- An explicit output declining to produce a treat or no-treat decision because the evidence or the operating conditions are insufficient. Abstention is a designed result, not an error and not a silent default.
- In depth:Uncertainty and abstention
- Decision record
- The durable, machine-readable artefact retained for each decision: source assets, rights status, capture context, model and calibration versions, policy version, uncertainty, review history, outcome, and export. It is the object an audit reads.
- In depth:Decision traceability
- Decision traceability
- The ability to reconstruct after the fact exactly how a decision was produced, from source imagery through model, calibration, and policy to the exported output. Traceability is what makes a decision correctable as well as defensible.
- In depth:Decision traceability
- Operating envelope
- Also: operating conditions
- The stated conditions under which a model and policy have been evaluated — crop, pest, growth stage, region, capture geometry, light, and equipment. Results produced outside the envelope are flagged, not silently reported.
- In depth:Uncertainty and abstentionColorado potato beetle
- Threshold policy
- Also: biological threshold policy, versioned threshold policy
- The reviewable, versioned rule set that turns biological observations into a management decision: which life stages count, how density or defoliation is assessed, which crop stage applies, and what uncertainty is tolerated. Held separately from model code.
- In depth:Biological threshold policies
- Treatment decision
- An explicit output stating treat, no-treat, or abstain for a defined area of a field at a defined time, together with the evidence and policy version that produced it. A detection is not a treatment decision.
- In depth:Pest decision platformBiological threshold policies
- Uncertainty
- The recorded degree to which the inputs to a decision are unreliable or incomplete — image quality, ambiguous life stage, sparse coverage, or a model operating outside its evaluated conditions. Uncertainty is carried into the decision rather than discarded.
- In depth:Uncertainty and abstention
- Validation scope
- The precise boundary of what has been tested: the crop, the pest, the life stages, the regions, the capture systems, and the model and policy versions involved. A claim without a stated validation scope is not a usable claim.
- In depth:Colorado potato beetle
Field data
- Capture session
- A registered event in which imagery is collected: a known farm, field, crop, date, camera rig, calibration state, and capture geometry. Imagery that does not belong to a registered session lacks the context a decision needs.
- In depth:Field data verification
- Data rights
- The recorded permissions attached to a field asset — what it may be used for, by whom, and for how long. Training eligibility is intended to be an enforceable state in the platform rather than an undocumented understanding.
- In depth:Field data verification
- Dataset lineage
- The record of which assets, annotations, and adjudications composed a training or evaluation dataset, and which models were built from it. Lineage is what allows a suspect result to be traced back to its data.
- In depth:Decision traceabilityField data verification
- Image quality
- The measured suitability of an image for analysis: focus, motion blur, exposure, occlusion, resolution at the target, and coverage of the intended area. Quality is assessed against the task, not judged by eye.
- In depth:Field data verification
- Invalid input
- An asset that fails verification — missing rights, missing capture metadata, unknown calibration, or measured quality below the threshold for the task. An invalid input is excluded from decisions and training, not down-weighted.
- In depth:Field data verification
- Provenance
- Also: field-image provenance
- The verifiable origin and handling history of an image: where and when it was captured, by what equipment, under what agreement, and what has happened to it since. Provenance is established at capture, not reconstructed later.
- In depth:Field data verificationDecision traceability
Biology and pest management
- Colorado potato beetle
- Also: CPB, Leptinotarsa decemlineata
- A defoliating beetle of potato and other solanaceous crops whose egg masses, larval instars, and adults are visually distinct. Widely documented insecticide resistance is why its management emphasizes scouting and thresholds.
- In depth:Colorado potato beetle
- Crop consultant
- Also: independent crop adviser, agronomic consultant
- An independent adviser who scouts fields and recommends management actions to a grower. Often the person who actually shapes a treatment decision, which makes them distinct from both the grower and an extension specialist — and a distinct audience for a decision platform.
- In depth:Pest decision platformField data verification
- Crop scouting
- Structured in-field inspection of a crop to estimate pest presence, life stage, and pressure, typically by sampling plants along a route. Scouting is skilled, sampled, and time-limited — which is why its coverage and repeatability vary.
- In depth:Agricultural computer visionField data verification
- Defoliation
- Loss of leaf area caused by feeding damage, commonly assessed as a percentage of the canopy. Several pest programs express thresholds in defoliation rather than pest counts.
- In depth:Colorado potato beetleBiological threshold policies
- Degree-day accumulation
- Also: growing degree days, GDD, biofix, heat units
- Tracking insect development by accumulating heat units above a species-specific lower developmental threshold, counted from a biofix — an observed biological starting event such as the first adult seen in spring. Programmes use it to anticipate when a life stage will appear rather than waiting to find it.
- In depth:Colorado potato beetleBiological threshold policies
- Economic threshold
- Also: action threshold, treatment threshold, spray threshold
- The pest level at which intervention is considered justified because expected crop loss outweighs the cost of acting. Economic thresholds are set by agronomic authorities and local programs, not by a software vendor.
- In depth:Biological threshold policies
- Insecticide
- A regulated product used to control insect pests, applied under the terms of its label. Swathmark does not recommend, select, rate, or interpret the label of any insecticide.
- Integrated pest management
- Also: IPM
- A management approach that combines monitoring, biological and cultural controls, and threshold-based intervention so that treatment is applied in response to observed pest pressure rather than on a fixed calendar.
- In depth:Biological threshold policies
- Life stage
- Also: instar, first generation, second generation
- The developmental phase of an insect — for example egg mass, larval instar, pupa, or adult. Life stage changes both how a pest appears in imagery and how a management program treats it, so detecting the stage matters as much as detecting the species.
- In depth:Colorado potato beetle
- Pest pressure
- The intensity of pest activity affecting a crop over an area and period, expressed through counts, density, or damage rather than a single sighting. Pressure is a field-scale property; a detection is an image-scale observation.
- In depth:Biological threshold policies
- Resistance management
- Also: mode-of-action rotation, insecticide resistance management, IRM
- Deliberately varying control tactics, including rotating insecticide modes of action, so a pest population is not repeatedly exposed to the same selection pressure. It is central to Colorado potato beetle management because of the pest's widely documented resistance to multiple insecticide classes.
- In depth:Colorado potato beetleBiological threshold policies
- Specialty crop
- Fruits, vegetables, tree nuts, and horticultural crops, as distinct from broadacre commodity row crops. Specialty crops carry high per-acre value and pest-specific management, which is where targeted decisions matter most.
Output and equipment
- OEM integration
- The defined software interface through which an equipment manufacturer consumes a decision: geometry and coordinate conventions, the edge and cloud boundary, export formats, and the operating limits attached to the decision.
- In depth:OEM integration
- Precision agriculture
- The practice of varying field operations by location instead of treating a field uniformly, using positioning, sensing, and machine control. Site-specific pest treatment is a precision-agriculture application that depends on a trustworthy decision, not only on capable machinery.
- In depth:Pest decision platformOEM integration
- Spot spraying
- Also: targeted application, site-specific treatment
- Applying product only to the locations that require it rather than across a whole field. Spot spraying is an equipment capability; whether a given location requires treatment is a decision question.
- In depth:OEM integrationPest decision platform
- Treatment zone
- Also: machine-readable treatment zone
- A georeferenced area expressing where a treat decision applies, together with the decision reference, policy version, and validity window. It is a representation of a decision, not an instruction to a machine.
- In depth:OEM integrationDecision traceability
Topics that use these terms
- Agricultural computer visionWhat agricultural computer vision estimates from field imagery, what it cannot establish on its own, and why a detection is not yet a treatment decision.
- Pest decision platformHow a pest decision platform differs from pest detection: the verified inputs it requires, the treat, no-treat, or abstain outputs it produces, and who uses it.
- Decision traceabilityWhat decision traceability means in practice: the provenance, model and policy versions, review history, and decision records needed to reconstruct a treatment decision.
- Biological threshold policiesWhat a biological threshold policy contains, how it differs from a detection, who sets thresholds, and why policy is versioned separately from model code.
- Uncertainty and abstentionWhy an agricultural decision system must be able to abstain, when it should, how uncertainty differs from abstention, and what happens after an abstain result.
- Field data verificationThe capture metadata, rights, calibration, and image-quality checks a field image must pass before it can support a decision or enter a training dataset.
- OEM integrationHow an equipment manufacturer would consume a Swathmark decision: interface design, the edge and cloud boundary, treatment-zone exports, and compatibility status.
- Colorado potato beetleColorado potato beetle biology and life stages, what computer vision can detect for each, why Swathmark starts here, and the current validation and evidence status.
Swathmark does not provide agronomic, pesticide-label, or treatment advice. Definitions here are descriptive. They do not state a threshold value, recommend a product, or interpret a pesticide label.