The Challenge:

Our client wanted to understand whether the application of machine learning to data (images, video) collected by an autonomous device could be used automate routine inspections currently only performed by humans. The solution needed to be scalable and extendable, integrate easily with our client’s existing tech stack, have a high level of confidence, provide easily digestible outputs, and provide a means to notify colleagues of required interventions.

Our Approach:

Our solution leveraged image data, technological expertise and computer vision algorithms to demonstrate the business applications and contextual value of data collected by autonomous devices. We developed an object detection algorithm allowing identification of a range of objects commonly found on industrial sites. We further developed an algorithm to detect equipment in an anomalous state. The outputs of the algorithms are consumed through a dashboard, which also functions as a messaging system alerting users about anomalies detected. The solution has been productionised in the cloud where the collected data is stored and processed, with the predictions displayed in the dashboard.

The Results:

By harnessing technological power and computer vision algorithms we have enabled autonomous inspection devices to observe and report on events during routine inspection missions. The benefits of autonomous inspection include improved safety, consistent data capture allowing better predictive outcomes, and improved efficiency through enhanced decision support leading to reductions in operational costs.