Recognising images with machine learning
This simple experiment uses machine learning to identify objects in uploaded images.
​
The page gives a very basic explanation for how computers can classify images using machine learning.
You will need
Smartphone
All you need is a phone with a camera
Other objects to detect
This can be a large variety of objects
Try the experiment!
We have made a special separate page for this activity, which contains the image classifier. Read some of the information below, and head over to the lab page when you are ready.
​
Please note, it is best to use this page over WiFi to avoid using mobile data. No information is sent by your device using this page, so images taken remain private and on your device.
Image recognition using machine learning
Image classification
Image classification is where a computer is able to recognise which objects appear in an image. This is used in many applications, some of which you may use every day.
​
For example, what if you want to use a computer to recognise whether a chair is present in an image? This is easy for humans, but very difficult for computers, because they do not have the knowledge to understand what a chair looks like. There are also many variations of different types of chairs, so we need to ensure that the computer understands the general properties of what a chair looks like.
​
For a long time, this was considered a very difficult task, but this is made much easier through the use of deep-learning.
Deep learning for image recognition
Deep learning is a type of machine learning which allows us to automatically identify the presence of objects in images. These work by using filters to identify the type of general features which might be present in an image of an object.
​
If you think of different examples of chairs, you will realise that they all have certain things in common; they generally possess similar shapes and textural features that allow us to recognise them as chairs. By taking thousands of images of chairs, and showing them to our computer, we can allow the computer to 'learn' some of the common characteristics of a chair, which will help the computer to identify chairs in general.
​
These tasks generally use many thousands of 'training' images to learn the general features of an object. However, once the program is trained, the computer should be able to effectively identify chairs in an image.
Where is image classification useful?
Of course, it is not just images of chairs that we can recognise. Image classification has many uses, some of which you may use every day.
The camera application in your phone using image classification to identify faces, which helps them focus on the correct part of the scene. Self-driving cars use image classification to identify objects around them, allowing them to drive safely. Image classification can also be used in healthcare, for instance to classify whether an illness is present based on an image of a cell.
