Computer Vision

Computer vision is the branch of artificial intelligence that allows computers to interpret and make sense of visual information -- photographs, video frames, scans, and live camera feeds. Rather than seeing an image as a grid of coloured pixels, a computer vision system is trained to recognise objects, scenes, faces, text, and actions within that visual data. It is one of the most commercially active areas of AI research, and it underpins technology most people encounter without realising it.

The field advanced dramatically once deep learning became practical in the early 2010s. Before that, computer vision relied on hand-crafted rules that were brittle and limited. Deep learning allowed models to learn features directly from millions of labelled images, producing systems that outperform previous approaches on almost every visual task.

How it works#

At the core of most computer vision systems is a type of neural network called a convolutional neural network (CNN). These networks process an image in layers. Early layers detect simple features -- edges, corners, gradients of colour. Later layers combine those features into more complex patterns: curves become shapes, shapes become objects. By the time information reaches the final layer, the network has built up a rich representation of what is in the image.

Training a computer vision model requires large quantities of labelled images. For an object detection model to reliably identify cars, it must be trained on thousands of images where cars have been marked with bounding boxes and labelled. The model learns to associate the visual patterns of cars with that label, and generalises to new images it has never seen. The quality and diversity of the training data is a major factor in how well the system performs in the real world.

Examples#

Supermarket self-checkouts. When you hold an item in front of a self-checkout camera, computer vision identifies what the product is -- whether it is a loose apple or a branded bottle -- and adds the right price to your basket. The system has been trained on thousands of images of different items from many angles and under different lighting conditions. Some retailers are now extending this to fully camera-monitored aisles that log what you pick up and charge you automatically as you leave.

Medical imaging. Radiologists use computer vision tools to help review scans for signs of disease. A model trained on thousands of annotated X-rays or MRI scans can flag areas that match patterns associated with tumours, fractures, or other conditions, drawing the clinician's attention to the most important regions. It does not replace the clinician's judgement, but it can act as a reliable second pair of eyes and help manage large volumes of scans more efficiently.

Why it matters#

Computer vision is already woven into daily life in ways that often go unnoticed: the Face ID on your phone, the speed cameras on motorways, the filters in social media apps, the accessibility features that read out image content for visually impaired users. As cameras become cheaper and AI processing costs fall, the range of applications is expanding rapidly.

For businesses, computer vision offers the ability to automate visual inspection tasks that would otherwise require trained staff. Quality control on a production line, monitoring workplace safety, analysing retail footfall patterns -- all of these are now accessible to organisations of a wide range of sizes.

To build a broader understanding of how AI systems learn from data, explore what is AI or browse the full learning paths to find a course suited to your level.