Field Of Vision

What it does

There are over 180 million visually impaired people worldwide who can only experience sport through audio. Our handheld device enables users to experience sport in a haptic format, allowing the user to feel the position of the ball on the pitch in real time.

Your inspiration

It was a viral video that inspired our journey, it was one of many viral videos of a man bringing his son to a game and tracing the position of the ball on a small board. However we saw this and thought that not everyone has a friend as great as this.. This is why we started Field Of Vision, for the people who didn't have a friend to do this for them.

How it works

We use artificial intelligence and computer vision to extract all the important information in real-time using the camera feeds from the game. We feed this information back to our device, which moves a magnetic finger-piece on the top of the device, mirroring the exact movements of the ball across the pitch. We’re also working on a new artificial intelligence method to track 3D ball position using a single, high quality camera. This will give us a very unique sell point in being able to offer our services by using existing broadcast cameras, without any of the set-up costs.

Design process

After creating a local network of future customers with whom we work with closely, we established what their requirements were for such a device. Based on this information, we developed a QFD Matrix which allowed us to evaluate which areas of the product’s development to allocate most of our effort and resources to. This steered us to focus on engineering a product with reliability and consistency, while forgoing things that were not as important as they may have initially seemed, such as portability and versatility. The prototype we are currently working on is aimed to enter service at live matches in stadiums. We designed our own Computer Vision model FootballNet to track the ball in real time. Instead of having to use anchors and sliding boxes like traditional object detection models, because we know the ball’s size beforehand, we treat object detection as a binary pixel-level segmentation task, classifying each pixel in an image as either the ball or the background. We trained the model on the publicly available Soccer dataset. It has 420k parameters and can track the ball to an accuracy of 85% at a speed of 150FPS on an Nvidia GPU.

How it is different

There are a few companies, backed by big names such as Santander, working on similar solutions. However, they have no AI-powered computer vision model, meaning they cannot scale and cover hundreds of live matches playing simultaneously worldwide. We are also the only product that is fully wireless and portable which is what the visually impaired community wants.

Future plans

We recently started a pilot programme with one of the largest football clubs in Ireland, Bohemians, and will begin testing our device there this summer. Several Premier League clubs and the Irish national team have shown considerable interest in purchasing our technology, after the pilot. After our success in Bohemians this summer we will begin our first paid games in the premier league in time for the 2021 season. This is to prepare for our short term goal of the Fifa World Cup, here in Qatar. After this, we will be expanding into other sports, and releasing a consumer product for use by millions at home

Awards

Enterprise Ireland's Student Entrepreneurs of the Year Award 2021 Qatar Sports Tech Accelerator Cohort 4 alumni Engineers Ireland Applied Student Project Award (undergraduate) Alsessor AI Accelerator Cohort 1 alumni Tangent Launchbox Accelerator 2021 participants Pioneers (February '21 winners, pioneer.app)