Currently, industrial sites are increasing production efficiency by adopting vision AI systems in various industrial applications, such as intelligent machines, cameras, and robots.

Texas Instruments (TI) announced the acceleration of smart vision product development with TDA4 processors and E-Con cameras through the 'TI Tech Day: Smart Vision AI and Communication Infrastructure' webinar on the 11th.

Manisha Agrawal, TI Product Marketing Manager, and Dilip Kumar, E-Con Systems Industrial Business Unit Product Manager, introduced ways to accelerate next-generation vision AI design. The design know-how for the 'TDA4 processor,' which can be used to realize next-generation smart machines, has been revealed.

Smart machines and cameras are key future technologies, bringing innovative advancements to various applications such as logistics warehouses, factories, last-mile delivery, services, smart agriculture, construction machinery, and smart traffic.

Manager Manisha said, “There is a need to add radar and LiDAR sensors to these smart machines to achieve better recognition performance in 2D/3D environments,” adding, “While more vision AI and 3D recognition processing are required, low-power processing to reduce battery load is essential for all tasks due to the limited form factor design.”

Therefore, solutions that satisfy development conditions requiring high performance, low power, and cost-effective design, and can accelerate product development in a rapidly changing development environment, are gaining popularity in the development field. Through collaboration with Econ Systems, TI provides a solution that can procure and immediately supply SOMs and camera modules.

Manager Manisha emphasized, “TI has a rich community to accelerate the processing and development environments of AI applications and Linux frameworks.”

In particular, the TDA4 processor introduced this time features a scalable dual-core processor based on the Arm Cortex-A72, support for up to 12-channel camera connections, as well as a hardware vision accelerator and an AI accelerator. Accelerators provide solutions for computationally intensive industries such as vision processing, ISP, bit lens distortion correction, scaling, and noise filtering.

A platform is in operation that provides solutions to developers, enabling AI development in no-code or low-code using the 'Edge Impulse' tool, and allowing the use of machine learning algorithms on edge devices by also providing AI tools necessary for rapid development.

TI provides 'TI Model Zoo' as a tool to optimize these trained AI models, offering optimization capabilities for representative models such as classification, detection, semantic segmentation, pose estimation, and face recognition. In addition, it includes popular latest models such as ResNet, RegNet, and the Yolo X series.

Dilip, a manager at Econ System, stated, “Through cameras compatible with the TDA4 processor, development can be accelerated in various application fields such as robots, mobile devices, and smart traffic control cameras.” Econ System, which has shipped 200,000 cameras this year, is contributing to the expansion of the vision AI market.

Meanwhile, TI operates an education academy to revitalize the related ecosystem and share developers' product development know-how, providing development technology insights.