A newly published study explores how artificial intelligence could make freshness assessment of white button mushrooms faster, cheaper and much more accessible.
Mushrooms are highly perishable, making reliable quality assessment important throughout packing, storage, transport and retail. Advanced techniques such as hyperspectral imaging can detect subtle changes in mushroom tissue, but these systems are relatively expensive and difficult to deploy throughout the supply chain.
Researchers have now developed an approach that transfers the knowledge obtained from hyperspectral imaging to a much smaller AI model that works with ordinary RGB images, the type produced by a smartphone camera.
The system was trained to recognise physical deterioration patterns in Agaricus bisporus. Once this knowledge had been transferred to the lightweight smartphone model, it achieved a classification accuracy of 81.11% while requiring only 0.2 million parameters.
Perhaps more interesting for practical use is its speed. The researchers report an inference time of just 18.6 milliseconds on a standard mobile CPU, allowing freshness assessments to be performed in real time without specialised computing equipment.
The technology could eventually give growers, packers, distributors and retailers an inexpensive additional tool for monitoring mushroom quality. Instead of relying only on visual judgement or laboratory equipment, a smartphone or similar device could potentially provide an instant AI-assisted freshness assessment.
The researchers also developed the system to cope with the more complex backgrounds found in retail environments, suggesting that the idea is already being considered beyond controlled laboratory conditions.
Further real-world validation will be necessary before such technology becomes part of everyday quality control. Nevertheless, transferring sophisticated spectral knowledge into something as accessible as a smartphone illustrates how advanced AI could gradually find its way into very practical parts of the mushroom supply chain.
