The goal of this assignment is to develop vision-based systems to prevent delamination during the 3D printing process. Cameras mounted inside the printer will capture images of the products from different angles, while intelligent algorithms, based on machine learning, will be trained on these images to detect delamination (also known as the “spaghetti” problem). Based on the detection results, the printing process will be aborted or continued. The project will explore recent approaches in machine learning including semi-supervised learning, Bayesian generative models, variational inference and data augmentation.
Signify will provide the means and methods (read hardware) to obtain the required image data, based on the requirements as defined by TU/e. The company will make a first base set of data, containing examples of the required image data, available at the start of the assignment (scheduled for February 2020).
Before the start of the project in February 2020, the candidate will:
- study the use case in collaboration with the company and the state of the art in literature,
- help the company defining the requirements for the collection of image data.
Supervision: Loek Tonnaer, Mike Holenderski, Vlado Menkovski