Posted on: September 16, 2020 at 10:12 AM    

We often hear the term artificial intelligence thrown around, a phrase which can be interpreted in many different ways. Artificial intelligence has undoubtedly changed how we interact with technology, from facial recognition on our smart phones, to Netflix recommendations. Most of us interact with artificial intelligence every day without even realising it. It is easy to think that these tools are reserved for the likes of Google or Apple, but these techniques and technologies are becoming more and more accessible across a range of industries and applications.

A particularly exciting field of artificial intelligence is "computer vision", which involves teaching a computer to understand imagery, allowing identification of objects within an image or what the image itself represents. It can sound like a deceptively simple task when we perceive the world so effortlessly. Computers lack the intuition that we have for such tasks.

Supported by Callaghan Innovation, I've joined Abley's Digital Engineering team as a Data Science Researcher for the next six months. My role is to research and develop machine learning based routines for automated processing and extraction of features from dense spatial data sets, and to explore how we could apply these new techniques to improve road safety outcomes.

So far, my time at Abley has been focused on understanding and defining the current need for tools and solutions. Leveraging the rich technical knowledge within the team, we’ve compiled a range of new and exciting ideas for how AI could be applied. Using a vast array of data sets, from still photos to video footage and LIDAR generated point clouds, our goal is to take this wealth of information and apply it differently. Could we use it to detect signs along a corridor and where they may be missing? Will it help us understand the condition of our roading assets without needing to put people’s safety at risk through on-site manual assessments? Can we develop processes that identify hazards before they become a reality? From these ideas, I’ve spent time understanding the existing research and complexity of each idea. Not all computer vision problems are created equal.

This is the first of a series of blogs. I'll be keeping you up-to-date on the progress we make and attempt to demystify the world of computer vision and machine learning!

Blog written by Joe Duncan, Data Science Researcher

Joe image 0 NOXoMg3xUgZUw388