Slide 1: Making edge AI a reality
Making edge AI a reality Products & enabling technology to accelerate edge AI adoption
Slide 2: How intelligent things are evolving
How intelligent things are evolving Smart Connected & secure Transformative Autonomous decision making Multimodal connectivity – Seamless user experience enabling high bandwidth to and Energy efficient from the cloud Changing work & lifestyle Personalization Identify, authenticate & protect Enhancin
Slide 3: What is artificial intelligence
What is artificial intelligence Perception The ability of machines/objects to Reasoning perform tasks that typically require a Learning human intelligence. Decision making From a technical perspective, AI is another way to program a machine, enabling it to extract meaningful information from data. 3
Slide 4: Using AI algorithms implies two phases
Using AI algorithms implies two phases Training Inference Model deployment AI models are produced using historical Inference is the process of using a trained machine datasets and a training engine/framework. learning model to make predictions or decisions based on new, unseen data. Dataset Training
Slide 5: The explosion of AI-enabled devices is accelerating the inference shift from the cloud to the tiny edge
The explosion of AI-enabled devices is accelerating the inference shift from the cloud to the tiny edge Inference in the cloud Inference on the device Inference at the edge Cloud-centric On device-centric Tiny edge-centric Enabled by a different class of hardware and software. Making AI more sustain
Slide 6: Inferring at the edge brings substantial benefits
Inferring at the edge brings substantial benefits Ultralow latency 01 Reduced data transmission Real-time applications 10 to generate meaningful information Enhanced privacy and security No data sharing in the cloud Sustainable on energy Low data, low power Lower cost of inference to enable a new cl
Slide 7: Opening a new range of embedded applications
Opening a new range of embedded applications Upgrade existing devices Reduce bom by shifting from Offer advanced services with AI-based services mpus to mcus without cloud costs Arc fault People Object detection detection classification Predictive Sound Pose & gesture maintenance analysis estimation
Slide 8: Edge AI on microcontrollers and smart sensors is a key enabler for smart applications
Edge AI on microcontrollers and smart sensors is a key enabler for smart applications Battery management Arc fault detection Face & object recognition Cities Gesture & motion recognition Home appliances Industry 5.0 & buildings Safety | Access control Environmental monitoring Anomaly detection Predi
Slide 9: 10+ hours
AR-Glasses ~15-40% 10 hours Energy saving per On single-charge washing cycle Virtual sensor Tire pressure measured through the e-motor current consumption Predictive Computer vision maintenance on MCU Multisensors and Cost-effective, compact, learning on device and low-power solution Electrical arc
Slide 10: Explore inspiring case studies
Explore inspiring case studies Gain valuable insights from real-world case studies on how to improve your embedded solutions using edge AI. Edge AI case studies Edge AI case studies 10
Slide 11: How ST is accelerating edge AI adoption
How ST is accelerating edge AI adoption
Slide 12: 10+ years of research, development, and deployment
10+ years of research, development, and deployment ST Neural-ART accelerator in STM32N6 ST Edge AI Suite Research using Edge AI on ST Edge AI ML techniques for Automotive MCUs Developer Cloud visual search Edge AI for sensors (MLC) Introduction of AutoML Edge AI devices 2012 2015 2018 2019 2020 2021
Slide 13: Taking up the edge AI developer's challenge
Taking up the edge AI developer’s challenge Hardware Performance Software Create, deploy Security & maintain ML models Power efficiency 13
Slide 14: A comprehensive approach to help developers accelerate their product transformation
A comprehensive approach to help developers accelerate their product transformation AI ST Edge AI Suite STM32 MCUs STM32N6 MCU AI Hardware Software platforms tools STM32 MPUs STM32MP2 MPU AI AI Stellar MCUs SPC5 MCUs MEMS with MLC and ISPU Click on the products to find out more AI With embedded AI a
Slide 15: Bringing advanced compute capabilities to enable a new class of embedded machine learning applications
Bringing advanced compute capabilities to enable a new class of embedded machine learning applications AI Sensors enabled with a machine learning core or an intelligent sensor processing unit (ISPU) for in-sensor AI. STM32N6 with proprietary ST Neural- ART Accelerator for unmatched levels of perform
Slide 16: First version of Neural-ART Accelerator embedded inside STM32N6
First version of Neural-ART Accelerator embedded inside STM32N6 ST proprietary neural processing unit (NPU) ML performance uplift* Specialized hardware accelerator performs advanced AI inference Achieving edge AI performance within MCU power budget “The most powerful accelerator seen in this class o
Slide 17: In-sensor AI for real-time processing: The intelligent sensor processing unit
In-sensor AI for real-time processing The intelligent sensor processing unit MEMS sensor 3-axis accelerometer & 3-axis gyroscope Raw data from MEMS sensor Sensor hub to collect data from additional external sensors (up to 4) ISPU processes data locally (a few µW) ISPU core Only meaningful data Stand
Slide 18: In-sensor AI for real-time processing: The machine learning core (MLC)
In-sensor AI for real-time processing The machine learning core (MLC) MLC is an in-sensor classification engine based on decision tree logic Sensor with machine learning core Motion sensing Main Sensor data Computation block Decision tree processor Accelerometer Filters Metaclassifier External Gyros
Slide 19: Enabling products and software
Enabling products and software
Slide 20: STM32 MCU offering for running AI
STM32 MCU offering for running AI SENSING (TIME SERIES) AUDIO VISION Enabling major edge AI technologies HORSE + DOG + CAT SENSING SENSING Software AUDIO tools for any VISION user profile User-friendly AutoML tool AI model benchmark, optimizer for STM32 MCUs and code generation for STM32 MCUs Large
Slide 21: STM32 MPU offering for running AI
STM32 MPU offering for running AI SENSING (TIME SERIES) AUDIO VISION Enabling major edge AI technologies HORSE + DOG + CAT SENSING Software AUDIO tools for any VISION user profile X-LINUX-AI A complete AI framework for STM32 MPUs Large choice of general purpose & accelerated hardware 21 STM32MP1 & S
Slide 22: STM32 is the de facto platform for embedded AI
STM32 is the de facto platform for embedded AI ML Commons #1 General purpose ML Perf Tiny benchmark MCU provider Multiyear top contributor in edge AI benchmarks Only online platform to offer MCU benchmarking 22
Slide 23: Sensor for running edge AI
Sensor for running edge AI Software tools • Profile and optimize NN and ML models Create end-to-end to develop • Configure the MEMS sensor-to-cloud in-sensor AI machine learning core solutions features and ISPU A broad range of evaluation & prototyping tools 20+ smart sensors released 6-axis IMUs 6-
Slide 24: Automotive MCUs running AI
Automotive MCUs running AI Edge AI provides more safety, better efficiency and car maintenance, and a personalized driving experience. StellarStudioAI Flexibility to import generated libraries into more complex application-specific projects AI SPC5-STUDIO-AI Easily validate and characterize the conv
Slide 25: A comprehensive approach to help developers accelerate their product transformation
A comprehensive approach to help developers accelerate their product transformation ST Edge AI Suite STM32 MCUs STM32N6 MCU Hardware Software platforms tools STM32 MPUs STM32MP2 MPU Stellar MCUs SPC5 MCUs MEMS with MLC and ISPU Click on the products to find out more 25
Slide 26: Creating an edge AI solution requires multiple skills and tasks
Creating an edge AI solution requires multiple skills and tasks Embedded ML Engineer Embedded Software Engineer Data scientist Hardware Engineer "I need to focus on essentials. "I mainly work in a scripting "I need simple tools to test my AI I prefer to start from existing environment for dataset so
Slide 27: 50+ case studies
By bringing solutions to engineers and data scientists at every stage of their development, the ST Edge AI Suite accelerates edge AI adoption. 50+ case studies 10+ free software tools Unified AI core technology st.com/st-edge-ai-suite 27
Slide 28: Access to the enabling hardware and software for running edge AI on MCUs, MPUs, and smart sensors
Access to the enabling hardware and software for running edge AI on MCUs, MPUs, and smart sensors. Empowering developers, data scientists, and product designers to harness the power of edge AI. A digital experience available in English, Chinese, and Japanese. 28
Slide 29: Free tools to run edge AI on MCUs, MPUs, and smart sensors
Free tools to run edge AI on MCUs, MPUs, and smart sensors Find the tools you need to optimize and deploy machine learning algorithms, from data collection to final deployment on hardware. Bring your Bring your OR own data own model AutoML Model Ideation Data Benchmark Deploy End-to-end flow optimiz
Slide 30: Find out more at st.com/edge-ai
Find out more at st.com/edge-ai © STMicroelectronics - All rights reserved. ST logo is a trademark or a registered trademark of STMicroelectronics International NV or its affiliates in the EU and/or other countries. www.st.com/trademarks For additional information about ST trademarks, please refer t