This week, BMW published what’s being referred to as SORDI (Synthetic Object Recognition Dataset for Industries), which is described as the largest open-source dataset for AI applications in production. Conceived with the purpose of streamlining the AI learning process and to strengthen no-code AI, SORDI features more than 800,000 photorealistic images of everything one might find in a given production facility, and divides them into 80 separate categories. The photos are of conventional items and equipment such as storage bins, pallets, forklifts, and other “objects of particular relevance to the core technologies of automotive engineering and logistics,” according to BMW. The dataset was created by BMW with partners Microsoft and Nvidia, both of which the automaker works closely with to streamline production. Idealworks, BMW’s logistics innovation spinoff, also played a role.
“The BMW Group has been using artificial intelligence since 2019. AI has already been utilised in various quality assurance applications in production at the plants. SORDI, the new, synthetic dataset makes AI models much faster to train and AI considerably more cost-efficient in production,” says Michele Melchiorre, senior vice president of BMW Group production system, planning, tool and plant engineering.
The dataset was built with Nvidia’s Omniverse platform, which BMW is using to plan factory floor layouts as efficiently as possible. Photos of objects and labels are uploaded by way of a rendering pipeline from BMW’s Tech Office in Munich, and are then synthesized in high-definition photorealistic quality so say can be used in the creation of AI models. BMW says SORDI can be used by both IT professionals, “to develop and tailor AI solutions for manufacturing,” and by production employees to, “maintain mature AI systems for validation purposes ready for the start of production.”
BMW hopes to democratize artificial intelligence, and SORDI is freely available to software developers from the automaker’s presence on GitHub. Other items of note include the BMW Labeling Tool Lite and accompanying AI training tools, which allow users to manipulate AI, even if they lack required technological experience, to get the most out of what SORDI offers in terms of the acceleration and training of AI models in production applications.—Alex Tock
[Photos courtesy BMW AG.]