Agri-food product characteristics change with time. These characteristics could change due to factors related to ageing, during storage or lying in our kitchens. Beyond maturity, factors like environmental conditions during plant growth, nutrient input, pathogens, storage conditions etc. greatly impact product characteristics. Also new varieties of products differ in their characteristics. As more and more companies rely on automation, they depend heavily on decisions (fruit is ready to be picked, ornamentals are of the desired quality, vegetables are disease free etc.) made by machine/deep learning (ML/DL) and machine vision methods (the eyes and brain behind the hardware). Current practice is the use of static ML/DL models. If, however, the distribution of inputs changes, because of changing product characteristics, such static models become invalid and lead to poor predictions.
Our goal with this project is to develop DL methods which can adapt to changing product characteristics and refine models as new data becomes available. The challenges addressed in the project are a direct result of the current industry demands and reflected in the sector specific use-cases. Overall, this project is contributing towards the next generation automation solutions which provide adaptive and precision production and food processing