“How can visual sensor data be analysed to improve the detection of defects and failures on wood surfaces for the automotive sector?”
ABOUT THE TECHNICAL CHALLENGE
For our technical challenge we are looking for people with various backgrounds – from analytics and informatics to design and data visualisation – who can improve failure detection on wooden surfaces through analysing visual data and visualizing these results.
As one of the most promising materials for the automotive and aerospace sector, wood has several unique properties, such as low thermal conduction, excellent vibration absorption, low material cost, and above all a raw material, that can be sustainably sourced within Austria and Europe and could potentially contribute to more value creation and resource independence of one of Europe’s key industries.
However, due to its natural origin, certain “defects and impurities” like branches, cracks, resin, markers, etc. of the material occur regularly and must be detected and extracted before it can be processed.
Things you should consider
- Create and train your own algorithm
- Our challenge partner will provide a data set containing 43.000 labelled surface defects (images) and a total of 10 different failure categories
- There will be wooden test boards provided to test and evaluate your algorithm
- The partner will provide additional data for different types of wood to train a Convolutional Neuronal Network
- Visualize your results through dashboards, graphic components, etc.
- GPUs will be made available to you
Evaluation criteria & questions
- The most important evaluation criteria will be the accuracy of your algorithm – hence, how well can you detect the surface defects?
- How are the actionable insights visualized for the user?
- Can the algorithm be applied to other types of wood?
- What other use cases/industries could the solution be applied to?
Does that sound interesting?
Apply now for one of the limited participation spots.