Extending machine learning to industrial IoT applications at the edge with AWS
summary
Claim: edge ML reduces latency and scales IIoT by moving inference on-device and retraining in cloud. Prototype: a Toradex module with an NXP i.MX8M and MIPI camera runs object detection on a mini conveyor (pasta demo) using TensorFlow MobileNet locally, with model labeling, training and optimization in SageMaker, deployment and offline runtime via AWS Greengrass, and telemetry over IoT Core/IoT Analytics to cloud dashboards. Two industrial pilots include a paper mill digital twin and SKF bearing-as-a-service.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 1:55Industry 40 trends
- 5:30Industrial IoT use cases
- 10:47Computer vision
- 17:04Virtual cycle
- 18:38Analytics
- 19:46Reference architectures
- 21:08Captive data
- 22:12Industrial data lake
- 24:08Industrial IoT reference architectures
- 24:42Customer success stories
- 28:22Wrapup
- 29:37Solution Repository
- 30:50Device Catalog
- 31:39Outro
related talks
49mEnable IoT with Edge Computing and Machine Learning
52mMachine Learning in IoT Applications
24mUsing AWS IoT Greengrass to make inferences at the edge
31mBuild Your Intelligent Edge
31mBring Cloud Connectivity and Intelligence to the IoT Edge
27mFrom Edge to Enterprise: Intelligent IoT Architectures
29mLife on the Edge
14mUnlock the power of IoT: Intro to NTT’s edge analytics platform