 Okay. Hello guys. Thanks for joining this AI sessions. So my name is Yvonne Rannou. I'm in charge of technical marketing for Germany and some other key customers. I'm located in Munich and I'm covering essentially Bavaria and some other customers. Okay. So let's start on this AI artificial intelligence. So I will start with a small marketing traditions. After that, you will go with Guillaume a little bit more in depth in terms of technical, what is behind this big world AI. And we'll spend 60 minutes together to practice on the current use case with the AI Cube tools. So artificial intelligence, we launched a new tools. The name is STM2 Cube AI. So this is an extension of the Cube, you know, very well. And we launched these tools at Embedded World this year. So it's now fully available and we are supporting customers with AI application with these tools. So this neural network, to build an application of a neural network, you have to go different steps. The first one is data capture. And for this, you need some hardware with some sensor. So this is a first break you need. After that, when you have a big collection of data, for sure you need to make some cleaning and you need to build your neural network. So this is not a part where we are directly involved because this is not our business and we have, there are a lot of parties, third parties, offering tools to do this. Like Keras, like Lazag and some other guys. So you have to use existing tools on the market to build your neural network on your workstations. After that, you have to train your neural network for your applications. And again, this work has to be done on the workstations. Until you have your model on your neural network working as you expect, you have to convert it to the Embedded World. And this is where ST is bringing innovations with these new tools. This cube AI converter. So what is it? It's a converter from a neural network working on the workstations with a lot of power, a lot of resources. Optimize it for the STM32. Okay, and this is what we will do together. We'll practice to see the added value of these tools. And when you have your neural network working on the STM32, you can start to do some process on analyzing for your applications. So what ST is providing? There are the different steps, the data capture, the build the neural network. So we're providing some tools to help and to facilitate these activities. So you have some hardware, but also some apps on your smartphone to configure the hardware and to offer something that is of use for these data collections. After that, you will have to use the cube AI neural network converter. This is what we'll do together. And finally, you will have your application running on STM32 with your artificial intelligence. So here, this is one example of the sensor tile. So I think you are familiar with this very small integrated hardware. We'll make a demo of this with Guillaume. So this is one example of the ready to use hardware from ST. That could be one bricks for this AI applications. Another one is IoT node. So this is a board you have in front of you. So this board is also very nice hardware ready to use. We have chosen this one because it has a ST link integrated. So it means it will be easier to flash and to make some debug on this board. This board has a lot of sensors, a lot of connectivity as well. So what else in terms of collecting data? So we offer the whole solutions starting from the hardware to collect all the data to offer you some apps on your mobile phone to configure the data. And finally, we are working on developing a huge partner ecosystem solution around the AI. Why? Because this AI is moving applications. Every month, we are seeing some new players and we have to follow. So this is really a strategic activity today in ST to follow the trend and to put some innovations with the STM32. So today, when a customer is coming and say, okay, I'm making an initial sensor, conditional monitoring and I need to put more intelligence on it. Okay, so what we do is we recommend some partners to help how to build a neural network. And until they have done these steps, ST could support them for the conversions of the neural network into the STM32. So there are different steps and we have to work with partners because we will not support the whole part for AI. This is too much. Okay, so here, this is some example of neural network, so very well-known name like Keras. So this is the one we will use today. But there are also Lasagne, Café and much more. There are many different neural network models that are existing on the market today and it's growing and growing. Here, in the middle, this is a CUBE AI. So this is a converter ST is offering and ST is supporting for these activities. And after that, for sure, we target STM32, so the STM32 Cortex-M but also the MP1 will come in the future as well. So in terms of front map, so today we support some neural network from the market, the floating point support. We offer different hardware for the data collections. We will put more features like the sensor flow light support, some new features with the quantizations, for example. Next year will come the ONIX support and some other features. So you see we have a complete front map in terms of tools and software implementations. What is coming next? Artificial intelligence at the sensor level, at the edge level is very nice but more and more you need some accelerator, hardware accelerator. So this is what we are considering for STM32 roadmap to integrate some hardware accelerator for AI. So it will come on the STM32 but it will come as well on the STM32 MP1, MPX roadmap to have some hardware accelerator to support a complete variety of AI applications from the sensor up to more advanced image recognition applications and so on. So ST wants to be present on this market and we have a complete roadmap in terms of tools and also in terms of hardware. So today we will demonstrate what we have on top of hardware plus the Qube AI converter. We have as well what we call some function pack. Some function pack is complete solutions to make some demonstrations. And we have one ready for the audio classification so we will demonstrate this but you will be able to in one click to demonstrate different environment you are with the STM32. So this is one function pack and the second one is human activity recognitions. So this is a complete solutions that we have today available. So we make some demos. So to summarize what ST is offering today for AI supporting. First of all the Qube AI so these are the tools we will show you today and we will practice together. Some hardware reference board like the sensor tile or the IoT nodes you have in front of you. Mobile application to help you to configure this board. Function pack this is completely application the full demo level. The community and the ST partners that we are working to reference more and more people working on artificial intelligence. The goal is to build a complete community around the STM32 for AI. So here is the summary of the different STM32 families that you are very well aware of it. The dark blue is to show all the families that are already supported by the STM32 Qube AI tools. So this is the majority of the families. The family where you see a light blue is Cortex M0 family. So this is not the first target for AI applications because you need generally some power processing but this is supported by partners. Or maybe if it makes sense we will support it in the future as well.