Identification of Shape Differences Among a Class of Materials
One example of this problem is the identification of strawberry shape. Strawberries that are conical are capable of being run through roller systems to align the berries in a way that they could be processed (e.g. topped by a water jet to remove the greens). Other berry shapes, such as wedges and rounds cannot be easily aligned by roller systems. Up to this point, however, it has not been possible to quickly identify one shape of strawberry versus another to enable their effective sorting during processing and packaging.
After considering a number of features that could potentially be used to discriminate between shape types, Appareo developed a deep neural network that used a combination of features that significantly outperformed the state of the art.