How Google uses pattern recognition

How Google uses pattern recognition to make sense of images

Computers don’t 'see' photos and videos in the same way that people do. When you look at a photo, you might see your best friend standing in front of her house. From a computer’s perspective, that same image is simply a load of data that it may interpret as shapes and information about colour values. While a computer won’t react like you do when you see that photo, a computer can be trained to recognise certain patterns of colour and shapes. For example, a computer might be trained to recognise the common patterns of shapes and colours that make up a digital image of a landscape such as a beach or an object like a car. This technology helps Google Photos organise your photos and lets users find any photo with a simple search.

A computer might also be trained to recognise the common patterns of shapes and colours that make up a digital image of a face. This process is known as face detection, and it’s the technology that helps Google to protect your privacy on services such as Street View, where computers try to detect and then blur the faces of any people that may have been standing on the street as the Street View car drove by.

If you get a little more advanced, the same pattern recognition technology that powers face detection can help a computer to understand characteristics of the face it has detected. For example, there might be certain patterns that suggest a face is smiling or has its eyes closed. Information like this can be used to help with features such as Google Photos’ suggestions of films and other effects created from your photos and videos.

Similar technology also powers the face grouping feature available in Google Photos in certain countries, which helps computers detect similar faces and group them together, making it easier for users to search and manage their photos. Read more about face grouping in the Google Photos Help Centre.

How Voice Search works

Voice Search allows you to provide a voice query to a Google Search client application on a device, instead of typing that query. It uses pattern recognition to transcribe spoken words to written text. We send the utterances to Google servers in order to recognise what was said by you.

For each voice query made to Voice Search, we store the language, the country and our system’s guess of what was said. We keep utterances to improve our services, including to train the system to better recognise the correct search query if you have given your consent to such data use. We do not send any utterances to Google unless you have indicated an intent to use the Voice Search function (for example, pressing the microphone icon in the quick search bar or in the virtual keyboard or saying “Google” when the quick search bar indicates that the Voice Search function is available).