This fleet safety software business was able to reduce cost and time with AI anomaly detection.



The Business

Based in South Africa, the business provides risk management consultancy and driver safety software to companies with large numbers of drivers. Their aim is to provide a full solution for the purposes of driver safety and they have a team of dedicated technologists focused on intelligence analytics, telematics, and fleet management.

The Problem

The company relies on IoT solutions that include cameras, tracking devices, and other sensors to monitor the safety of drivers in the fleets that they are contracted to manage. All of the data from these sensors is aggregated in a centralized platform.

The main issue that the business faced was predicting camera failure. A failed camera means that they are not capturing the data that they need to ensure proper driver safety. In the past, camera quality had been monitored by humans. When camera quality begins to fail, this means that the business needs to replace that camera. When cameras malfunction, agents alert drivers to let them know it needs to be fixed or replaced.

It’s a massive investment in time, and can be an unreliable method of actually spotting discrepancies due to the large numbers of images that staff members review on a daily basis. The business is constantly adding new fleets to their operations, and some projects can have more than 10,000 units to monitor, which means an increase in time and personnel is necessary with each new contact.

They sought a means of scaling the monitoring process and quality assurance while still being able to rely on quick response times and cost efficiency.

The Solution

The solution was quite simple. Stradigi AI created a model within Kepler that would detect when the quality of the picture was suspect, or when cameras were down completely. Images were added in bulk to a machine learning model that spotted the anomalies. Once the anomalies were detected, the business was able to relay that information to the appropriate department so that they could change out or repair those cameras.

The Results

The image processing is now done without any human intervention, leading to a dramatic decrease in the need for human resources for this task that now can be applied elsewhere.

Response times for repairing cameras are now much faster, which improves driver safety and has led to a marked improvement in cost efficiency.


> Anonymous


> South Africa


> Fleet Safety

Workflow used

> Image classification with anomaly detection

This site uses cookies to provide the best experience. By continuing to browse the site, you are agreeing to our use of cookies.