Master thesis project: Deep learning for label inspection

We are proposing a thesis project that focuses on implementing label inspection using neural networks and deep learning.


The Eagle Vision Basic Scout system can inspect labels and artworks on misprints, damages like dents and scratches, and other label damages and differences. This is done using "traditional" computer vision and image processing techniques. A recent trend however is demand for higher inspection performance, and better human-like classification of differences, distinguishing between relevant differences caused by a damage or misprint, and accidental differences that are not relevant for the inspection. The expectation is that using new machine learning techniques, especially deep-learning, will make this better automatic classification possible.

The focus of the master thesis project will be on developing a framework that can implement the described classification. This will be done using the best machine learning stack for the job. So for example using Tensorflow, Pytorch or any other suitable library. The inspection systems themselves are built in C++. An extension of the product, could be integrating the machine learning framework in the inspection system software.

Your skills

  • Bachelor in a relevant subject like computer science
  • Knowledge of computer vision
  • Knowledge of Machine Learning
  • Knowledge and some experience with the state of the art Machine Learning stack

Ga terug