Projects facts

  • Scope
    Identifying health and living condition problems across hundreds of thousands of insects at various stages of their lifecycle.
  • Period
    2020
  • Solution
    An automated visual inspection solution that uses AI to accelerate issue detection and manual intervention.

Results

  • Increased insect protein yield

  • Automated visual inspection processes

  • Improved process efficiency and living conditions

The challenge

Successfully breeding thousands of creatures

Even on small farms, keeping a close eye on animal health, wellbeing and living conditions can be tough. But when you’re breeding and rearing hundreds of thousands of creatures at once, it’s impossible to do manually.

That’s the challenge our client faced. They breed insects on an industrial scale for use as a sustainable protein source. To optimise the harvest of this protein, it is needed to optimise the circumstances for the insects.

With such a large volume of creatures to monitor, manual inspection simply wasn’t an option. So, they began exploring ways to apply AI across their operations – with the help of Royal HaskoningDHV Digital.

The approach

Enabling automated visual monitoring

To understand how AI could be applied to help solve the client’s challenge, we first needed to understand what the visual indicators of insect health and living conditions are. It turned out that experts are looking for food residues, dead larvae, deviations in insects and secretions when monitoring the insects. Even for an expert it is not easy to spot these deviations in boxes containing 100,000 insects.

After working with internal experts to map those out deviations, work began devising a video-based AI solution that would use cameras to continuously generate visual data of all the farm’s insects. Then, we constructed AI models to process that data and automatically detect anomalies that require attention.

With so many insects to monitor, and small time-windows to take action against some health issues, the accuracy and speed of the model were of the utmost importance.

We were also able to automate some direct interventions (like differentiated feeding based on the analysis), helping to resolve them within seconds of the issue being detected, before any further negative impacts are caused.

The impact

Improving efficiency and protein yield

With our visual AI solution in place, inspection of all insects is now automated – helping to identify health problems earlier, accelerate direct interventions, and improve total usable protein yield.

The solution has also been an important learning tool for the client. As a pioneer in the sustainable insect protein field, they’re constantly working to improve the efficiency and quality of their operations. By producing and analysing vast quantities of visual data, their solution supports that aim directly.

The end result is a more efficient farm, that can output more protein at a lower cost – while gaining invaluable insights that will aid future growth and improvements in insect breeding practices.

With the visual AI solution in place, inspection of all insects is now automated – helping to identify problems earlier, accelerate interventions, and improve total yield.