1st VDI Conference on “Machine Vision” with a presentation by ID Engineering

The first voluntary sustainability report from ID is now available online on the homepage. Voluntary because ID is not yet legally obliged to report. According to the current EU directive on sustainability reporting (CSRD – Corporate Sustainability Reporting Directive), companies currently only have to report if they meet at least two of the following three criteria: more than 250 employees, a balance sheet total of more than 25 million euros or net sales of more than 50 million euros.

Although there is a proposal at EU level to raise this threshold to 1,000 employees in the future – it has not yet been decided and is therefore not legally binding. Because a sustainability strategy gives many companies the opportunity to position themselves in a future-proof and sustainable manner, ID has taken the initiative.

The roles were perfectly divided in tandem. Dr. Wanner first explained the scientific level, what synthetic data is and how to produce it in a sensor-realistic way. ID Machine Vision specialist Sartor then presented practical examples of its importance for industrial image processing.

In principle, the use of synthetic image data in industry is always advantageous where it is not possible or simply too expensive to generate or collect errors in order to use them for training AI or deep learning. With a high variety of products or a rapid change of products in the automation chain, synthetically generated data can increase quality assurance.

Example 1: Seat inspection in the automotive industry. Here, cracks, scars, seam defects or dents can be simulated by synthetically generated surfaces in order to train the AI in such a way that exactly these defects are detected directly from the start of production without having previously found and trained them in reality.

Example 2: Pot inspection in the household industry. Deep learning with synthetic data is used, for example, to detect color defects, enamel scratches, faulty printing or a lack of gloss on surfaces and processed in the production software.