Computer Vision in Practice: See, Sort and Inspect with Machines

A coach that teaches by asking — not by lecturing

The Lighthouse Coach is an AI Socratic coach built on Cognosco Learning's Computer Vision in Practice: See, Sort and Inspect with Machines course, in the AI at Work category. It does not play videos at you. It holds a conversation: it asks you a question, listens to your answer, names the misconception it hears, affirms what you got right, and asks the next question — the way a great coach works in the room.

A vision system fails first in the framing, not in the model. That is this course's premise. Most failed vision projects did not fail because the model was weak. They failed because nobody had said, precisely, what job the seeing was for; because the images were collected from a tidy corner of a messy process; because the labels encoded three different opinions about what counted as a defect; or because the threshold was left at whatever value the demo shipped with, as if a miss and a false alarm cost the same.

It starts with the five shapes. Every vision job, stripped of its setting, is one of five shapes — classify, detect, track, measure, read — and the course's deciding move is a single discipline: name the shape, name the decision the shape feeds, and name what the shape cannot tell you, before any other project conversation. At Dunvarra Packaging Co., the course's running company, one vague sentence — "We want the cameras to check the cartons" — hides three different jobs with three different data needs and three different error prices, and the course teaches you to split it before anyone collects an image or prices a camera.

It works in two movements. First, Module 1 — Frame It, Feed It, Price the Errors — work through Concepts 1–9, finishing with Vision on Line 2; then, Module 2 — Run It in the Real World — work through Concepts 10–18, finishing with The First Season.

Both capstones produce real artifacts — a Project Plan for putting vision inspection on Dunvarra's Line 2, which runs at 240 cartons a minute and ships about 38,000 cases a week: the reframed request, the site sheet surveyed to its worst honest hour, the collection and labeling design, the buy, adapt, or build decision, and the priced threshold with its evaluation protocol; and an Operating Plan for running that system through its first season — deployment inside its timing budgets, a drift watch with diagnosis before response, a rework-bench review loop with real override authority, a classified failure review, and the owner's one-page state report.

Two boundaries are held throughout. This course is non-clinical, without exception: every application is inspection, sorting, counting, document capture, or monitoring of spaces and goods — no medical imaging, no diagnosis, no clinical decision support anywhere in it. And the course owns the project, not the machinery underneath it: it never teaches network architecture and never asks you to write model code, so where a model family or a vendor tool is named, it is named at judgment level, with specifics flagged to verify as current, because model names and tool features change faster than judgment does.

The course is self-paced and always on — roughly 30 minutes per concept, about 9 hours in full, taken in as many sittings as you like. You can type your answers, or talk out loud: the Voice Coach speaks its questions and listens to your spoken replies. Learners who already know the material can test out concept by concept.

Computer Vision in Practice: See, Sort and Inspect with Machines is part of The Lighthouse Signature Series — the online counterpart to the one-day, in-person learning Cognosco Learning delivers.