ST opens up its edge AI tool to other chips
STMicroelectronics has developed an edge AI tool that works across all its smart sensors, microcontrollers and microprocessors and is even opening up the tool to chips from other semiconductor companies that use the ARM Cortex-M core.
STMicroelectronics has negotiated a special license to offer its edge AI tool for framework creation, development and deployment for microcontrollers from competing chip makers.
The ST Edge AI Suite has a common optimizer for all ST devices, supporting ARM cores and hardware accelerators such as the STM32N6 and the STM32MP2. A unified command line interface (CLI) pulls together the different AI tools in the suite, which will be free for ST parts.
- ST to launch its first neural microcontroller with NPU
- Industry 4.0 edge AI powered microprocessors
- Code free machine learning tool for edge IoT equipment
“We have decided to make the deployment free for STM32 devices and to make it available for all and any Cortex-M MCUs under a special license agreement to offer it as widely as possible to accelerate the adoption of edge AI,” said Remi el-Ouazzane, president of microcontrollers and digital ICs at STMicroelectronics.
“With the Nano v4.3 we are going to refer to all the Cortex-M devices in the tool, so when a developer triggers a project the ML library will be Cortex- M specific without hardware acceleration and so those libraries can be compiled on any implementation. We did not want the customers to feel they were locked into us and we wanted to remove any friction for the adoption,” he said.
“It is not free of charge but it allows customers to have a unified environment internally.”
- ST launches dedicated microcontroller edge AI cloud farm
- STMicroelectronics acquires Cartesiam for edge AI tool
The tool suite also allows frameworks to be re-trained with data from the embedded system. This will require additional tools such as a data logger to capture and annotate the data.
“We are developing a data logger that is a data logger generator not only for capture of data but also for annotation so the ability to retrain with data from the field will be easier with a common data logger across the end points. We have other ideas for a closer loop between a retraining and the model accuracy but it is not there yet today.”
The Edge AI Suite is already being used for consumer AI such as washing machines, and in automotive on the Stellar microcontrollers to use edge AI algorithms to create a virtual temperature sensor in electric vehicle motors.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
