Texas Instruments
Vision processor for scalable edge AI performance
TI offers a family of vision processors based on Arm Cortex consisting of six modules. These make it possible to equip applications such as machine vision applications and autonomous mobile robots with extensive image and AI processing functions.
The family, which consists of the 'AM62A', 'AM68A' and 'AM69A' processors, is supported by open source tools for evaluation and model development. This is complemented by uniform software that can be programmed via industry-standard APIs (Application Programming Interfaces) as well as frameworks and models. The platform of vision processors, software and tools helps to develop and scale edge AI designs for a wide range of systems.
TI 's vision processors are based on a system-on-chip (SoC) architecture with a high degree of integration. The integrated components include an Arm Cortex-A53 or Cortex-A72 CPU, a third-generation TI image signal processor, memory, interfaces and hardware accelerators that achieve an AI processing performance of 1 to 32 TOPS (Teraoperations Per Second) for deep learning applications.
The vision processors in the family are as follows:
- The AM62A3, AM62A3-Q1, AM62A7 and AM62A7-Q1 versions support one or two cameras with less than 2 W in applications such as door intercoms or intelligent systems for retail.
- The AM68A supports between one and eight cameras, for example in machine vision applications. It offers AI processing performance of up to 8 TOPS for video evaluation.
- The AM69A has an AI processing performance of 32 TOPS for one to twelve cameras. It is designed for 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 with a public beta version of TI's free, open-source Edge AI Studio tool. The web-based tool allows for quick and easy development and testing of AI models.










