The detection of defects during cutting tool fabrication poses a great challenge for automated inspection systems due to high variation of surface textures, defect appearance, and overall lack of training data.
In this work, you will work with us in close collaboration with the Fraunhofer Society to develop new rendering algorithms that are capable of rapidly generating such training data.
In particular, your work will involve improving current sampling techniques (i.e, path guiding, environment-map sampling), as well as effective descriptions of (the lighting of) inspection setups with the ultimate goal of rapidly generating large amounts of realistic training data.
If you are interested in contributing to research in this context, please contact Hiroyuki Sakai.
Requirements
- Motivation to contribute to research in an interdisciplinary team
- Willingness to deal with theory and mathematics