NMI在NM180100上使用Ambiq Apollo3 SoC实现本地AI推理,仅需1.35秒!

Описание к видео NMI在NM180100上使用Ambiq Apollo3 SoC实现本地AI推理,仅需1.35秒!

在这段视频中,Northern Mechatronics(NMI)——一家总部位于加拿大的物联网设备射频和天线设计领域的领导者,以不到2秒(准确地说是1.35秒)的时间在其旗舰产品NM180100上运行了机器学习(ML)推理模型,出色地展示了可选字符识别的性能,这一切得益于Ambiq的Apollo3 SoC提供支持。

NMI证明了使用合适的组件在本地运行边缘和端点计算推理的可能性,包括动作识别、声学特征识别和基于Tensorflow Lite实现的神经网络架构等应用。
NM180100是一款支持LoRa® 和Bluetooth® 5的低功耗模块,提供turnkey解决方案,拥有充足的存储空间和内置处理器,可实现无缝集成。Ambiq的Apollo3 SoC是一款超低功耗、高度集成的微控制器平台,基于其专有的亚阈值功率优化技术(SPOT®)提供增强的turboSPOT®性能。

欲了解更多信息,请访问NMI网站 https://www.northernmechatronics.com/...
和Ambiq网站 https://ambiq.com/apollo3/

In this video, Northern Mechatronics (NMI) a leader in RF and antenna design for IoT devices based in Canada, delivered an outstanding performance of less than 2 seconds (1.35 seconds, to be exact) of running the Machine Learning (ML) inferencing model for optional character recognition on its flagship NM180100, enabled by Ambiq’s Apollo3 SoC.

NMI demonstrated the possibility of running inferencing LOCALLY for edge and endpoint compute with the right components for applications, including activity recognition, acoustic signature recognition, and neural net architecture via Tensorflow Lite, and more.

The NM180100 is a LoRa® Bluetooth® 5 Low Energy Module, a turnkey solution with long range, ample storage, and a built-in processor for seamless integration. Ambiq’s Apollo3 SoC is an ultra-low power, highly integrated microcontroller platform supercharged with its turboSPOT®, based on the patented Subthreshold Power Optimized Technology (SPOT®.) Visit NMI https://www.northernmechatronics.com/... for more information and Ambiq website https://ambiq.com/apollo3/

Follow us:
• LinkedIn:   / ambiqmicro  
• Facebook:   / ambiqmicro  
• Instagram:   / ambiq_ai  
• Twitter:   / ambiq_ai  
• Blogs: https://ambiq.com/blog/

About Us: https://ambiq.com/

Timestamp:
0:00 Intro
13:45 Outro

#NMI #Ambiq #Apollo3 #SoC #AI #EdgeAI #LocalInference #FastAI #TechInnovation

NMI在NM180100上使用Ambiq Apollo3 SoC实现本地AI推理,仅需1.35秒!

Комментарии

Информация по комментариям в разработке