Computer Vision in Production and Retail: online course with internship agreement
Computer Vision in Production and Retail teaches you to specify, validate and run camera inspection on production lines and in warehouses. It covers lighting and pixels per millimetre, rules versus learned models, defect datasets, thresholds and drift, barcode and dimensioning tunnels, and the EU Machinery Regulation and AI Act. You study 100 % online for 180 days, with an internship agreement included, for €200.
- 180 days of access
- 100 % online, at your own pace
- Level: beginner · No prior machine vision experience required
- Assessment: 10 quizzes (50 questions) and a final project
- Internship agreement included
- Certificate with a verifiable QR code
- Languages: Spanish, English, French
- Price: €200
Who it is for
- Quality and production staff who want to automate visual inspection on a line
- Technicians who want to install, commission and maintain vision systems in factories or warehouses
- People who have used image-analysis software but never learned optics, lighting or resolution
- Engineers preparing vision equipment that will be built into machines sold in the EU
What do you need to start?
Prior knowledge
- No prior knowledge of machine vision or machine learning is needed
- Comfort with simple arithmetic on paper: products, sums, percentages and rates
- Willingness to read datasheets, specifications and regulatory texts carefully
Software and equipment
- A computer and an internet connection
- No manufacturer's vision software is required: the course compares Cognex, Keyence, SICK, Omron and HALCON through specifications, not menus
- Public research defect-image collections can be used for practice; check their licence, which often allows non-commercial use only
What you'll be able to do
- Calculate whether a camera resolves a defect by working out the pixels across it on the part
- Decide, with reasons, whether an inspection task needs rule-based vision or a learned model
- Choose between classification, detection and segmentation according to the answer the factory needs
- Build a confusion matrix and compute recall, precision, accuracy and reject rate for an inspection cell
- Set the decision threshold that minimises the combined cost of escapes and false rejects
- Write a vision specification and compare two supplier offers on cycle time and five-year cost
- Diagnose barcode no-reads from contrast, quiet zone, module width and print grade
- Determine whether a camera on a machine needs a notified body or falls under the AI Act's high-risk regime
Skills you will practise
- Lighting and optics
- Pixels per millimetre
- Convolution and CNNs
- Defect dataset labelling
- Confusion matrix and thresholds
- Edge deployment and drift
- Barcode reading and OCR
- Parcel dimensioning
- Machinery Regulation (EU) 2023/1230
- AI Act and software licences
Syllabus
- Machine vision in the European factory: what the job actually is — Why manual visual inspection lets rare defects escape, and what a machine vision professional actually does in factories.
- Imaging fundamentals: optics, lighting, sensor and pixels per millimetre — How optics, lighting and sensor combine, and why pixels across the defect, not megapixels, decide whether a camera can see it.
- From image to decision: classical algorithms, convolution and learned models — Rules versus learned models, convolution worked by hand, classification versus detection versus segmentation, honest testing on other batches, and model licences.
- Datasets and labelling for defect detection: the rare-class problem — Which images to collect when defects are rare, how labelling defines a defect, and checking dataset licences before commercial use.
- Deployment and MLOps at the edge: drift, retraining and versioning — Confusion matrices, cost-based thresholds, edge versus server sizing, per-part logging, drift sentinels and a controlled retraining loop with rollback.
- Reading a vision spec instead of a brand: Cognex, Keyence, SICK, Omron, HALCON — Turning a line requirement into a written specification and comparing supplier offers on cycle-time arithmetic and five-year cost.
- Logistics and warehouse vision: barcodes, dimensioning and OCR — Reading tunnels, no-read causes, exposure and time budgets, OCR uncertainty, and parcel dimensioning as a legally controlled measurement.
- The Machinery Regulation (EU) 2023/1230 and your camera — What the Machinery Regulation requires, whom it binds, and how it frames a vision system built into a machine.
- Self-evolving safety functions: when you need a notified body — The two questions—continued learning after market placement and dependence of a safety function—that decide notified-body involvement under Annex I.
- After the Digital Omnibus: the AI Act timeline that actually applies — When a vision project is high risk under the AI Act, from which date, what applies otherwise, and copyleft licence checks.
Internship agreement
With this course, an intern could support a quality or production department by checking lighting and resolution for an inspection task, labelling defect images or calculating recall, precision and reject rates for an existing cell. In a warehouse or dispatch area, tasks could include analysing barcode no-reads and their cost in exceptions per shift. In a machine-building firm, an intern could help prepare vision specifications or compliance documentation.
The internship can run at the same time as the course, within the 180 days of access.
Frequently asked questions
What kind of work does Computer Vision in Production and Retail prepare me for?
Computer Vision in Production and Retail prepares you to specify, commission, validate and maintain camera systems for part inspection and for warehouse reading and dimensioning tunnels. The materials note that warehouses and parcel operators hire far more technicians for this than research labs hire researchers. The course does not turn you into a notified body or replace a manufacturer's own product training; it teaches you to reason across suppliers and regulations.
Do I need to master one brand, such as Cognex or Keyence, to work in machine vision?
The course argues otherwise. Vacancies list Cognex, Keyence, Omron, SICK or HALCON because factories install whatever solved their last problem. What carries from one employer to the next is the ability to write a vision specification, calculate pixels per millimetre and cycle time, and compare five-year costs. Menu structures, tool names and configuration formats stay behind with the licence you happened to use.
Which laws, standards and methods does the course actually cover?
It covers the Machinery Regulation (EU) 2023/1230, including when a self-evolving safety function requires a notified body under Annex I, and the AI Act timeline after the Digital Omnibus. It also covers legal metrological control of parcel dimensioning, copyleft and dataset licences, confusion matrices, cost-based thresholds, drift monitoring, retraining with rollback, convolution, and classification, detection and segmentation.
What could I do during the internship, and who finds the company?
You find the host company yourself; the school then issues the internship agreement and its annex, and the company signs electronically, usually within one to two working days. Suitable tasks include measuring an inspection cell's error rates, labelling defect images, diagnosing barcode no-reads in a dispatch area, or helping draft a vision specification or supporting compliance documents for a machine.
More Technology & AI courses
Content updated: 07/10/2026