Computer Vision Model Testing Services

Independent evaluation for object detection, classification, segmentation, tracking, OCR, face systems and vision-language models across real deployment conditions.

What We Test

SystemMeasurement
Object detectionPer-class precision, recall, localisation, FPPI and miss rate by size
ClassificationPer-class accuracy, confusion and calibration
SegmentationPer-class IoU, boundary accuracy and instance separation
Tracking & videoIdentity switches, fragmentation and re-identification
OCR & document AICharacter and field accuracy by script, layout and capture
Face & person systemsRates across cohorts, conditions and appearance factors
Vision-language modelsGrounding, counting and hallucination
3D & point cloudAccuracy by range, density, weather and sensor

Why mAP Is Not the Whole Story

An illustrative aggregate mAP of 0.91 can still contain an unusable night-rain, small-object or rare-class slice. All values are illustrative.

MetricQuestion answeredInterpretation
Per-class APWhich classes workReport instance counts
Precision / recall at operating thresholdProduction trade-offAvoid unused thresholds
FPPIOperator burdenAbsolute alerts matter
Miss rate by size / distanceWhere detection falls offTreat small objects separately
IoU distributionWhether boxes are usableA mean can hide a split
Per-slice breakdownCondition and cohort failuresWhere production failures live
Confidence calibrationWhether scores mean what logic assumesConfident errors are costly
ID switches & fragmentationIdentity continuityTracking averages can hide failure

Test Sets Built for Deployment

We control device, lighting, weather, geography, scale, occlusion, rarity and temporal drift. Image data collection can source missing evaluation imagery.

Distribution shift

  • Capture shift
  • Environmental shift
  • Semantic drift
  • Population shift

Face, Person and OCR Testing

Face and person systems are evaluated across cohort × condition cells. We use skin reflectance where it provides a more useful measurement than perceived skin tone. Formal fairness framing belongs in the AI Bias & Fairness Audit.

DimensionCoverageWhy it matters
ScriptLatin and eight major Indian scriptsConjuncts, diacritics and ligatures
Print qualityDigital to degraded or stamped-overCopy loss has its own curve
HandwritingPrint, cursive and mixedField accuracy diverges
LayoutTables, forms, rotation and skewStructure errors differ
CaptureScanner and phone imageryAngle, glare, curl and shadow
Field extractionCorrect value in correct fieldWhat workflows consume

Vision-Language Model Evaluation

We test visual hallucination, grounding, spatial reasoning, counting, text-in-image reading, fine-grained distinction and uncertainty.

Ground Truth Is Part of the Test

Reference annotations use documented edge cases, adjudication and reported agreement. Image annotation creates training labels; testing measures fitness to ship.

How an Engagement Runs

  1. Scope and slice definition
  2. Test-set design and sourcing
  3. Reference annotation
  4. Evaluation
  5. Failure analysis
  6. Report and walkthrough

What You Receive

DeliverableContents
Slice-level accuracy reportMetrics per class, condition, device, location and cohort
Failure atlasActual images grouped by cause and counted
Error taxonomyRoot causes ranked by frequency × cost
Operating-point analysisPrecision and recall at production thresholds
Versioned test setReusable imagery and reference annotations
Annotation agreementReference quality reported before findings
Class-definition findingsAmbiguities surfaced through review
Live walkthroughEvidence reviewed with the vision team

Where Vision Testing Goes Wrong

FailureConsequenceControl
Single mAP headlineA collapsing class disappearsPer-class and per-slice reporting
Training-like test setHome-turf measurementDeployment-led sourcing
Post-hoc slicesFlattering cutsPre-agreed slices
Noisy test annotationLabel error looks like model errorReference-standard adjudication
Unused thresholdsResults do not describe productionActual operating point
Small objects averaged inDistance failures vanishSize-band miss rate
Demographics aloneCondition effects stay hiddenCohort × condition
Character accuracy onlyFields remain unusableWorkflow-level metrics
Fluent VLM descriptionsHallucination scores wellGrounding and abstention
One-time evaluationDrift goes unseenVersioned re-testing

What We Do Not Do

We do not tune the client's model, report one unexplained aggregate, treat unaudited labels as ground truth, or support unlawful biometric surveillance.

How to Engage

ProgrammeBest fitWindow
Baseline vision evaluationFull test-set and slice measurement6–7 weeks
Test-set construction onlySourcing and reference annotation4–5 weeks
Release-cycle testingRepeat against baseline1–2 weeks
Deployment-site auditFocused drift investigation2–3 weeks

Related Services

Image Annotation Video Annotation Document & OCR Annotation 3D Point Cloud & LiDAR Annotation AI Bias & Fairness Audit Data Collection

Computer Vision Model Testing FAQs

What is computer vision model testing?

Structured measurement of how a vision model performs on imagery that reflects its deployment environment — reported per class, per condition, per device and per cohort rather than as a single aggregate figure.

Why isn't mAP enough?

Because it is an average, and production vision failures concentrate rather than spread. We report per class and per slice, at the confidence threshold you actually deploy.

Our model scores well but fails in production. Why?

Usually because the test set resembles the training data rather than the deployment environment. We examine capture, environmental, semantic and population shift.

Can you build the test images, or do we have to supply them?

We can build them across regions, devices, lighting and conditions when deployment imagery is missing.

Do you test face recognition for bias?

We test cohort × condition cells and appearance factors, including skin reflectance where it provides a more useful measurement than perceived skin tone. Formal fairness and regulatory mapping moves to our bias and fairness audit service.

Do you test OCR in Indian scripts?

Yes — Devanagari, Bengali, Tamil, Telugu, Gujarati, Gurmukhi, Odia and Urdu, across print quality, handwriting, layout and capture method.

Can you evaluate vision-language models?

Yes. We cover visual hallucination, grounding, spatial reasoning, counting, reading text in images, fine-grained distinction and abstention.

How good does our ground truth need to be?

Better than training data: documented edge cases, adjudication and reported agreement are required for a defensible test.

How is this different from your annotation services?

Annotation produces labelled training data. Computer vision testing produces error rates and failure analysis used to decide whether a model is fit to ship.

How long does it take, and is the test set reusable?

A first programme typically runs 6–7 weeks, or 4–5 with suitable imagery. The versioned test set is delivered for reuse.

Find the Failure Slices Your Average Is Hiding

Scope a Computer Vision Test Programme