Computer Vision: Where are we and what’s next for CV?

Computer Vision (CV) is an engine that can understand and separate the visual components of either a still or moving image. It currently has many practical applications but we are likely to see further uses in the next five years which will greatly impact and shape the tech sector.

The current uses of CV are wide ranging. From an everyday standpoint, companies such as Shutterstock and Facebook are using the technology to improve their user experience, by making photo navigation easier. At the other end of the spectrum of usage, surveillance and security agencies are also increasingly relying on CV technology in order to better analyse CCTV footage.

Image libraries employ their own proprietary algorithms to sift through thousands of images detecting faces and objects. The end result is that the user can find images that are visually similar to their initial request without having to rely, on what can be, limiting keyword data. Face-swap apps are currently a popular example of CV technology. Some such as MSQRD are attracting the attention of tech giants, with Facebook purchasing the start-up earlier this year.

The algorithms that underpin CV are constantly being refined via machine learning, therefore making what has been previously only possible, now practical. In Shutterstock’s case, CV was used to develop tools such as reverse image search and to offer users visually similar results. During the learning phase, Shutterstock’s engineers presented the computer with images that all had something in common, for example they contained a green apple. After some time, through trial and error, the computer learns what visual components make up a green apple and it adapts in order to present users with visually similar images of other green apples. The more images the computer sifts through and the more positive or negative feedback it receives, the more accurate the search results become.

The future

As a form of AI, the potential uses of CV are extremely far reaching.

This technology goes beyond helping us to access images that are visually similar to our initial request. Medical research is an area where CV can have a huge impact. Technology that can save lives is always going to be highly valued. The ability to analyse tissue samples to spot irregularities such as tumours is currently being honed. Institutions such as UCL are leaders in this space, using CV and image processing to make the analysis of MRI scans more efficient and accurate.

Those with disabilities and others who struggle with language are also being helped by the technology. The language of search and discovery is being shaped by CV. For Shutterstock specifically, we operate in over 150 countries around the world, and these developments help to break down language barriers. Tools like reverse image search, which allow users to search via images instead of keywords, will soon become more widespread. Similarly, Facebook’s recent development which can "read" photos and tell visually impaired people what appears in them is an exciting prospect for many.

The methods used by security and surveillance teams across the world are a further use of CV. Being able to pick the faces of wanted suspects out of a sea people is a useful tool in an age of increasing tension regarding national security, especially when crowded areas are the target.

One interesting innovation is anomaly detection which has the power to spot unusual or unexpected movement in crowds. In this case, CV is used in conjunction with other disciplines such as swarm theory and statistical sequential change detection. When used together, experts are able to identify what they term as “abnormal” events. This helps surveillance teams to cut through the noise and direct on-the-ground support to areas that are potentially experiencing some kind of threat.

Another area where CV is keeping us safe is in the development of autonomous cars. Without human input, driverless cars rely on CV in conjunction with radar, GPS and odometry in order to provide a safe and efficient form of transport for their users. CV is key here, as a driverless car will need to take extra care when in the vicinity of pedestrians or cyclists – these are nuances that often only CV can identify.

Alongside transport and safety, CV has the power to make life simpler for many of us. In the next five years our world will be shaped by many forms of AI, this is something we are already beginning to see the impact of – for example Google’s recent Go victory.

Personalisation is another key area that will increasingly employ the technology to provide a raft of services to consumers. For example, we may not be too far away from a time when a computer can analyse photos on your phone and begin suggesting pieces of artwork or furnishings that align with your tastes.

Kevin Lester, VP Engineering, Search and Discovery at Shutterstock

Image source: Shutterstock/agsandrew