Texas Instruments unveils six Arm Cortex-based vision processors: support for up to 12 cameras

The new family includes the AM62A, AM68A and AM69A processors, supported by open-source evaluation and model development tools, and general-purpose software programmable through industry-common application programming interfaces (APIs), frameworks and models. Using this platform of vision processors, software and tools, designers can easily develop and scale edge AI designs across multiple systems while reducing time-to-market.

When implementing vision processing and deep learning capabilities in low-power edge AI applications, Texas Instruments' new vision processors bring intelligence from the cloud to the real world by removing barriers of cost and design complexity.

These processors feature a system-on-chip (SoC) architecture that supports extensive integration. Integrated components include an Arm Cortex-A53 or Cortex-A72 central processing unit, third-generation Texas Instruments image signal processor, internal memory, interfaces, and hardware accelerators to deliver 1 teraflop to 32 teraflops for deep learning algorithms AI processing function of computing TOPS (IT Home Note: Tera Operations Per Second is a unit of processor computing power).

Vision processors in this family include:

AM62A3, AM62A3-Q1, AM62A7, and AM62A7-Q1, which can support one or two cameras less than 2W in applications such as doorbell cameras and smart retail systems. The family includes the AM62A3, the industry's lowest cost 1 TOPS vision processor.

The AM68A can support 1 to 8 cameras in applications such as machine vision, and up to 8 TOPS of AI processing for advanced video analytics.

The AM69A enables 32 TOPS of AI processing for 1 to 12 cameras in high-performance applications such as edge AI boxes, autonomous mobile robots, and traffic monitoring systems.

Beginning in the second quarter of 2023, designers can accelerate time-to-market for their edge AI applications through the public beta of Edge AI Studio, Texas Instruments’ free and open-source tool. This feature-rich, web-based tool allows users to quickly and easily develop and test AI models using self-created and TI-optimized models, which can also be retrained using custom data.