Project information
In this project, an existing AI model for potato defect detection has been further optimised. Previously, this model was used to identify one prominent anomaly per image of potato slices. The new task for the students was to detect multiple defects per image more accurately, thereby ensuring better potato quality.
What has been realised?
The project group carried out the project in several phases. First, they compiled an extensive dataset of potato slice photos, in which the different defects were labelled. Subsequently, they trained an AI model to automatically recognise these defects. Finally, an application was developed that applies the model to analyse potato slice photos and automatically determine which defects are present or when a slice is in good condition. This application can, in the future, be integrated into a machine for industrial application.
What has been learned?
The project group has learned how to define and describe an assignment with a company in a way that is both useful and actionable. In addition, they have learned how to constructively apply AI in an industry, where the system enhances existing processes without detrimental effects.
