An autonomously driving car using DL methods is creating a lot of rumor. Heterogenous CPU architectures - not only with GPU accelerators - are necessary. Is this suitable for machine vision, too? Of course, DL methods have the potential to solve machine vision applications, too. But does the sector have the army of engineers working in the automotive industry? No. Take a look at easy architectures and libraries. While the learning phase of DL needs huge (cloud-based) computing power, the application is much easier. Implemented on typical multi-core ARM or i-Core-based computer systems, DL systems can be realized. Programming, debugging - also important to be on the market in time. IMAGO's new VisionCam with a powerful Dual Core ARM Cortex-A15 or the flagship VisionBox LE MANS with an Octa Core ARM Cortex-A72 are coming with the needed computing power. Finally, not every new problem needs to be solved by DL - traditional methods with intelligent operators are still recommended.