fbpx

TensorFlow Lite for Microcontrollers (TFLM): Recent Developments


Date:  Wednesday, May 18 (Main Conference Day 2)

Start Time: 2:05 pm

End Time: 2:35 pm

TensorFlow Lite Micro (TFLM) is a generic inference framework designed to run TensorFlow models on digital signal processors (DSPs), microcontrollers and other embedded targets with small memory footprints and very low power usage. TFLM aims to be easily portable to various embedded targets from those running an RTOS to bare-metal code. TFLM leverages the model optimization tools from the TensorFlow ecosystem and has additional embedded-specific optimizations to reduce the memory footprint. TFLM also integrates with a number of community contributed optimized hardware-specific kernel implementations. In this talk, we review collaboration between BDTI and Google over the last year, including porting nearly two dozen operators from TensorFlow Lite to TFLM, creation of a separate Arduino examples repository, improved testing and documentation of both Arduino and Colab training examples and transitioning TFLM’s open-source CI framework to use GitHub Actions.

Track

Session Speakers

  • Advait Jain is a software engineer at Google who currently leads the TensorFlow Lite Micro team. He is a roboticist and software engineer with broad expertise across control systems, mechanics, sensing and perception, machine learning, systems-level thinking, and people, program, and product management. He works at the intersection of hardware and software and makes pragmatic design choices to build integrated systems with desired real-world task performance.

  • David Davis is a consulting embedded software engineer with many many years of experience in both consumer electronics and industrial applications. He started his career porting various flavors of the Unix operating system to mini- and micro-computers. This was followed by both embedded and applications level work for telephone switching systems, multimedia document editors, digital television and cable systems, digital video recorders, wireless music players, WiFi routers, and wafer test equipment. He is currently engaged in embedded development and support work involving TensorFlow on microcontrollers.

  • John Withers is a consulting automation and systems engineer with over 20 years experience in backend infrastructure, CI, and CD. For the last few years he has been working extensively in the embedded space doing work on containerization in low-resource environments, secure status monitoring and OTA updates for computer vision edge products, as well as CI pipelines for large ML codebases. Recent clients have included Nvidia and Google.


Share

TensorFlow Lite for Microcontrollers (TFLM): Recent Developments