 Hi, we are at TES 2020 and I welcome you in the AI lab. My name is Vincent Abreu and I work in the AI team as a computer vision expert. As you may know, last year, we released the STM32 Cube.AI tools that is able to convert pre-trained neural network into an optimized code for our STM32 products. This year, we are more focusing on what we could achieve with this tool, especially on computer vision, to address home appliance wearable industrial and toy markets. In this demo, based on STM32-H7 that has been done in collaboration with the CA French Research Institute, we are implementing face identification which is able to identify up to eight persons in a row. You can imagine to have this kind of application in home appliance for user, to increase user experience to change some settings parameters. This expression demo is based also on STM32-H7 and is able to detect your mode. Up to seven kind of expression is able to be detected like sadness, happiness. And you can imagine to have this kind of application on toys to increase user experience based on the user expression. Here we show the STM32 MP1 microprocessor on which we have implemented a TensorFlow Lite neural network for image classification. Image classification is able to classify picture among 1,000 classes and you can imagine to have this in a fridge or in a freezer to detect low supply. With a sensor tile box, we are demonstrating a distributed intelligence. One is done on the smart sensor for the human activity recognition and on the other part is done on the STM32 to do send audio classification. You can imagine to have this in a wearable device in order to detect that the user is driving to enable the BLE for example or to when the user is outside to disable the Wi-Fi in order to set some program. So this product is done by the company called 5Voxel. This is based on the 3D camera with a timer flight and is able to have a gesture recognition. You can imagine to have this kind of feature in the home appliance like extractor hood when you are cooking and your hands are dirty and you want to enable or disable the extractor hood. This demo is done by one of partners named Nalbi and they are providing a solution for face identification with a self-enrollment on which you can have up to 500 faces. So for the last demo, let me introduce you Mark from Cartesium that will show you what they are doing with the sensor tile. Hi, I am Mark Dupak, I am the founder of Cartesium. We are an ST Micro partner and we are based in France, US and Germany. So what we are doing here is software, machine learning software that runs on Cortex MCUs. I have a demo that I will show you today. So we have here a small sensor. The learning is made on the sensor as well as the analysis. So we decided to analyze a pretty difficult engine. You can see here a simulated fan. We actually have two fans. They go on, they go off, they stop, they go slow, they go fast, and all this is being random. So very difficult. So we have been doing the learning. So now this is monitoring the system and I will simulate a problem. So for example, if that's a fan here, air filter being dirty, so I'm fooling with the air and as you can see here on the small screen, the anomaly is being detected. So that's typically how it works. Very odd thing to analyze being done on the MCU. So in here, I have no time to do it today but obviously you can see more of it on www.cartisium.ai. With that's a software that will enable embedded developer to develop the solution that they will embed into their solution. So thank you very much for your time. If you want more detail, you can go on cartisium.ai or on st.com. Thank you very much. Goodbye.