BMW, the German car manufacturer, has developed an anonymization solution based on artificial intelligence. This model will prevent an object from being identified in a photo or video. The platform is based on the BMW Lite labelling tool, already designed by the brand.

Applications of artificial intelligence in BMW’s production system

As part of “Project AI”, the BMW Group had already published, in 2019, a first selection of algorithms and programs used in the production system, as Kai Demtröder, head of data transformation and AI for the brand, tells us:

“The huge amount of feedback we have received on the algorithms we have released has been very pleasing. Our BMW AI community is very pleased with the appreciation we are receiving from all over the world. We are seeing many more developments based on our source code and our code-free AI approach. We will continue to follow this approach and share our code-free AI algorithms so that AI can be made accessible to a wide range of users.”

One of the ways the automaker uses AI is to detect objects in production through image processing. BMW says its application offers a very good level of robustness, even under specific conditions where parameters may vary. AI-based image analysis, according to the brand, helps to maintain a higher quality of production. Another use of AI is in the development of autonomous intelligent robots for logistics tasks.

BWM has developed an AI-based anonymization solution that fits into the above-mentioned applications.

BMW’s anonymization solution

This software is mainly based on AI which is used to automatically classify certain areas of an image according to predefined parameters and characteristics. The system analyzes its areas and makes sure that all the areas that should become unrecognizable become so by blurring, masking or pixelating them. In the case of video, they can even be blocked or deleted so that they never appear.

The main technology used is the BMW Lite tagging tool, which allows users to classify photos with tags or train the AI model. Each label contains all the information contained in the image or in a portion of the video.

Markus Grüneisl, head of production system and digitization at BMW, said around the creation of this anonymizing software:

“AI applications support the quality control teams in tasks such as the inspection of parts and components, as well as in the development of our autonomous and intelligent logistics robots. The anonymizing algorithms that can now be used also guarantee optimal data privacy and information protection.”

The solution is called “No-Code AI”: this promotes the creation of codes by production staff to answer a particular problem, even if they are not trained in coding. This was one of the brand’s wishes, as Markus Grüneisl states:

“Making the anonymization solution intuitive to use was an important aspect of the development for us, to ensure that it can be easily used for a wide range of applications.”

The algorithms are also publicly available so that software developers around the world can modify and improve the source code if they wish.

Translated from BMW dévoile une solution d’anonymisation basée sur l’intelligence artificielle