Computer vision (CV) is the AI field teaching machines to extract meaning from images and video: what is in the picture, where the objects are, what happens. For a human "seeing" is natural; for a computer an image is just a grid of numbers (pixels). CV turns these numbers into understanding. This course gives a practical picture: what CV can do, how it works at the idea level, and how to apply ready tools — without deep math.
An image for a computer is numbers
A picture for a machine is a grid of pixels, each of which is numbers (brightness, color). The grid itself does not "say" what is on it. The CV task is to go from numbers to meaning: "this is a cat", "here is a face", "here is a defect on the part". It is non-trivial: the same object looks different under different light, angle, size — while a human recognizes it instantly.
What computer vision can do
What tasks does computer vision solve? Give examples.
Computer vision covers a whole spectrum of tasks — from simple to complex. The main types (which the course covers): 1) Image classification: determine WHAT is in the picture overall — one label per image. Example: "the photo is a cat / dog / bird", "this shot is a defect or normal", "the face mood is joy". Answers "what is it?". 2) Object detection: find WHERE the objects are in the image and what they are — boxes around each object with a label. Example: on a street photo outline all cars, people, signs. Answers "what and where?". Harder than classification — not one label but many objects with location. 3) Segmentation: determine which pixels EXACTLY belong to an object — a precise "mask" of the shape, not just a box. Example: outline the exact tumor contour on a scan, separate the road from the background pixel by pixel. Answers "which exact pixels are the object?". Even more precise than detection. 4) Face recognition: a special case — detect faces and/or determine whose face (identification). Example: face unlock, finding a person in photos. 5) Optical character recognition (OCR): extract TEXT from an image. Example: read text from a document photo, a number on a plate. 6) Assessment/analysis: counting objects, pose estimation, motion tracking in video, quality assessment. Where it is used (real-life examples): — Medicine: analyzing scans (X-ray, MRI) to help a doctor — in overview, from the AI-in-medicine course. — Manufacturing: quality control, defect finding on a conveyor. — Retail: product recognition, shelf analytics, scanner-less checkout. — Security/access: face recognition, video surveillance. — Transport: perception for self-driving cars (objects, signs, markings). — Agriculture: field analysis, plant condition by photo/drone. — Documents: OCR, paperwork processing automation. — Everyday: phone face unlock, photo-album search "show photos with a dog", filters/effects. The key idea: modern CV works on the same machine-learning and neural-network principles we covered (a model learns on many image examples rather than being programmed with "if the pixel is such" rules). It is learning on data that made CV powerful: previously people tried to write rules by hand (nearly impossible for real images), now a model learns to recognize from examples, as a child learns to recognize objects. The practical meaning for you: you do not need to be a neural-network or math expert to apply CV. There are ready models and services for classification, detection, OCR, face recognition. Understanding WHAT CV can do (task types) and how to apply it (ready tools, your data, quality evaluation), you can solve real tasks. The course gives this practical picture.
Learning on data, not rules
The key idea: modern CV works on the same machine-learning and neural-network principles we covered. A model learns on many examples of images rather than being programmed with "if the pixel is such" rules. Previously people tried to write rules by hand (nearly impossible for real images — too many variations of light, angle), now a model learns to recognize from examples, as a child learns to recognize objects. It is learning on data that made CV truly powerful.
You do not need deep math
The practical meaning: you do not need to be a neural-network expert to apply CV. There are ready models and services for classification, detection, OCR, face recognition. Understanding what CV can do (task types) and how to apply it (ready tools, your data, quality evaluation), you solve real tasks. The course gives this practical picture: CV task types, how recognition works at the idea level, applications and pitfalls — without diving into math.
Course rule: computer vision extracts meaning from images (classification, detection, segmentation, OCR etc.). It works through training neural networks on examples, not through hand rules. Apply ready tools — deep math is not needed.
Computer vision is a machines ability to extract meaning from images and video.
🧠 How does modern computer vision recognize images?