Charbax | Fraunhofer IIS embeddif.[ai] World's Smallest AI at Embedded World 2024 #ew24 @charbax | Uploaded April 2024 | Updated October 2024, 1 week ago.
The embedded world of 2024 showcases a groundbreaking development with the unveiling of the world's smallest AI unit, brought to life by Fraunhofer IIS's embeddif.ai. The event features an impressive presentation by Maximilian Kasper, who delves into the unique capabilities of this tiny AI solution. The star of the show is a compact computing unit capable of running complex AI models, demonstrating a new level of embedded artificial intelligence.
The small computing unit is based on an ARM Cortex M7 processor, a well-known architecture for embedded systems, usually seen in microcontrollers. This AI unit's impressive feature is its ability to run an AI model called Tiny YOLO, designed for real-time object detection. Kasper demonstrates how this small AI system can detect and identify people using a camera connected to the unit. Despite its tiny size, the system manages to process images at nearly 5 frames per second, with a resolution of 128x128 pixels, showcasing its high efficiency.
Another significant aspect of this AI unit is its efficient use of embedded C code, optimized for the Cortex M7. This allows the system to fully utilize the processor's capabilities, ensuring maximum performance. Alongside this demonstration, different setups showcase varied applications of the technology. One demo illustrates on-device training, where the same hardware setup can classify images of animals. The AI can detect a cow or a dog, displaying its predictions on a small screen. This adaptability opens up a range of possibilities for edge computing applications, where AI can operate without the need for a central server.
The flexibility to train on-device is one of the most innovative aspects of this tiny AI system. This is particularly useful for applications where a central server isn't feasible or real-time training is required. Kasper explains that the training data is supplied through a small SD card, which contains the ground truth information for several images. The on-device training uses backpropagation and stochastic gradient descent, all processed directly within the AI unit. This capability to perform training on the device distinguishes this technology from other AI solutions that require pre-trained models deployed from external sources.
Despite its size, the AI unit's power requirements are relatively high, which is expected given the complex computations involved in training and running AI models. During the event, Kasper demonstrated an untrained model, which often misclassified objects, showing that proper training is essential for accurate predictions. This underlines the importance of reliable training data and efficient learning algorithms for successful AI deployment in embedded systems.
In summary, the world's smallest AI unit by Fraunhofer IIS at Embedded World 2024 introduces a new realm of possibilities for embedded AI. With its compact design, powerful ARM Cortex M7 processor, and ability to run and train AI models on-device, this technology has the potential to revolutionize embedded applications. Its use of embedded C code, optimized for the processor, ensures high performance and efficiency, while the capacity to train on-device offers flexibility in various scenarios. The potential applications are vast, from smart sensors to real-time object detection, marking a significant leap in embedded AI technology.
Watch all my videos from the Embedded World 2024 here: youtube.com/playlist?list=PL7xXqJFxvYvjgUpdNMBkGzEWU6YVxR8Ga
Description by Chatgpt.
This video was filmed using the DJI Pocket 3 ($669 at amzn.to/4aMpKIC), watch all my DJI Pocket 3 videos here youtube.com/playlist?list=PL7xXqJFxvYvhDlWIAxm_pR9dp7ArSkhKK
Please support my work. You can donate any amount "Value for Value!" to my paypal.me/charbax or even better (no fees) to my wise account: charbax@gmail.com (you can join Wise for free at wise.com/invite/dic/nicolasc1941)
If you consider renting out a room or your home or your second home to make extra income, please use my link to sign up to Airbnb airbnb.com/r/ncharbonnier?s=6&t=061n0g I would get a €295 commission from Airbnb, when you list your space and complete a first qualifying stay. You will get €23, too. I appreciate this very much! I absolutely love Airbnb, it has changed the lives of millions of people around the world! Watch this video where I talk about Airbnbs in Turkey for example youtube.com/watch?v=g5rkFKThBew
Please Join as an Early Access Member on my YouTube channel at youtube.com/charbax/join and for about $5/month you get instant Early Access to all my videos immediately as I upload all my videos nightly from the conferences that I attend, instantly accessible to my members as I share them publicly at the rate of about 2 or 3 videos per day over the weeks following the show.
The embedded world of 2024 showcases a groundbreaking development with the unveiling of the world's smallest AI unit, brought to life by Fraunhofer IIS's embeddif.ai. The event features an impressive presentation by Maximilian Kasper, who delves into the unique capabilities of this tiny AI solution. The star of the show is a compact computing unit capable of running complex AI models, demonstrating a new level of embedded artificial intelligence.
The small computing unit is based on an ARM Cortex M7 processor, a well-known architecture for embedded systems, usually seen in microcontrollers. This AI unit's impressive feature is its ability to run an AI model called Tiny YOLO, designed for real-time object detection. Kasper demonstrates how this small AI system can detect and identify people using a camera connected to the unit. Despite its tiny size, the system manages to process images at nearly 5 frames per second, with a resolution of 128x128 pixels, showcasing its high efficiency.
Another significant aspect of this AI unit is its efficient use of embedded C code, optimized for the Cortex M7. This allows the system to fully utilize the processor's capabilities, ensuring maximum performance. Alongside this demonstration, different setups showcase varied applications of the technology. One demo illustrates on-device training, where the same hardware setup can classify images of animals. The AI can detect a cow or a dog, displaying its predictions on a small screen. This adaptability opens up a range of possibilities for edge computing applications, where AI can operate without the need for a central server.
The flexibility to train on-device is one of the most innovative aspects of this tiny AI system. This is particularly useful for applications where a central server isn't feasible or real-time training is required. Kasper explains that the training data is supplied through a small SD card, which contains the ground truth information for several images. The on-device training uses backpropagation and stochastic gradient descent, all processed directly within the AI unit. This capability to perform training on the device distinguishes this technology from other AI solutions that require pre-trained models deployed from external sources.
Despite its size, the AI unit's power requirements are relatively high, which is expected given the complex computations involved in training and running AI models. During the event, Kasper demonstrated an untrained model, which often misclassified objects, showing that proper training is essential for accurate predictions. This underlines the importance of reliable training data and efficient learning algorithms for successful AI deployment in embedded systems.
In summary, the world's smallest AI unit by Fraunhofer IIS at Embedded World 2024 introduces a new realm of possibilities for embedded AI. With its compact design, powerful ARM Cortex M7 processor, and ability to run and train AI models on-device, this technology has the potential to revolutionize embedded applications. Its use of embedded C code, optimized for the processor, ensures high performance and efficiency, while the capacity to train on-device offers flexibility in various scenarios. The potential applications are vast, from smart sensors to real-time object detection, marking a significant leap in embedded AI technology.
Watch all my videos from the Embedded World 2024 here: youtube.com/playlist?list=PL7xXqJFxvYvjgUpdNMBkGzEWU6YVxR8Ga
Description by Chatgpt.
This video was filmed using the DJI Pocket 3 ($669 at amzn.to/4aMpKIC), watch all my DJI Pocket 3 videos here youtube.com/playlist?list=PL7xXqJFxvYvhDlWIAxm_pR9dp7ArSkhKK
Please support my work. You can donate any amount "Value for Value!" to my paypal.me/charbax or even better (no fees) to my wise account: charbax@gmail.com (you can join Wise for free at wise.com/invite/dic/nicolasc1941)
If you consider renting out a room or your home or your second home to make extra income, please use my link to sign up to Airbnb airbnb.com/r/ncharbonnier?s=6&t=061n0g I would get a €295 commission from Airbnb, when you list your space and complete a first qualifying stay. You will get €23, too. I appreciate this very much! I absolutely love Airbnb, it has changed the lives of millions of people around the world! Watch this video where I talk about Airbnbs in Turkey for example youtube.com/watch?v=g5rkFKThBew
Please Join as an Early Access Member on my YouTube channel at youtube.com/charbax/join and for about $5/month you get instant Early Access to all my videos immediately as I upload all my videos nightly from the conferences that I attend, instantly accessible to my members as I share them publicly at the rate of about 2 or 3 videos per day over the weeks following the show.
![Fraunhofer IIS embeddif.[ai] Worlds Smallest AI at Embedded World 2024 #ew24](https://i.ytimg.com/vi/eSjFRkrNlOQ/hqdefault.jpg "Fraunhofer IIS embeddif.[ai] Worlds Smallest AI at Embedded World 2024 #ew24
The embedded world of 2024 showcases a groundbreaking development with the unveiling of the worlds smallest AI unit, brought to life by Fraunhofer IISs embeddif.ai. The event features an impressive presentation by Maximilian Kasper, who delves into the unique capabilities of this tiny AI solution. The star of the show is a compact computing unit capable of running complex AI models, demonstrating a new level of embedded artificial intelligence.
The small computing unit is based on an ARM Cortex M7 processor, a well-known architecture for embedded systems, usually seen in microcontrollers. This AI units impressive feature is its ability to run an AI model called Tiny YOLO, designed for real-time object detection. Kasper demonstrates how this small AI system can detect and identify people using a camera connected to the unit. Despite its tiny size, the system manages to process images at nearly 5 frames per second, with a resolution of 128x128 pixels, showcasing its high efficiency.
Another significant aspect of this AI unit is its efficient use of embedded C code, optimized for the Cortex M7. This allows the system to fully utilize the processors capabilities, ensuring maximum performance. Alongside this demonstration, different setups showcase varied applications of the technology. One demo illustrates on-device training, where the same hardware setup can classify images of animals. The AI can detect a cow or a dog, displaying its predictions on a small screen. This adaptability opens up a range of possibilities for edge computing applications, where AI can operate without the need for a central server.
The flexibility to train on-device is one of the most innovative aspects of this tiny AI system. This is particularly useful for applications where a central server isnt feasible or real-time training is required. Kasper explains that the training data is supplied through a small SD card, which contains the ground truth information for several images. The on-device training uses backpropagation and stochastic gradient descent, all processed directly within the AI unit. This capability to perform training on the device distinguishes this technology from other AI solutions that require pre-trained models deployed from external sources.
Despite its size, the AI units power requirements are relatively high, which is expected given the complex computations involved in training and running AI models. During the event, Kasper demonstrated an untrained model, which often misclassified objects, showing that proper training is essential for accurate predictions. This underlines the importance of reliable training data and efficient learning algorithms for successful AI deployment in embedded systems.
In summary, the worlds smallest AI unit by Fraunhofer IIS at Embedded World 2024 introduces a new realm of possibilities for embedded AI. With its compact design, powerful ARM Cortex M7 processor, and ability to run and train AI models on-device, this technology has the potential to revolutionize embedded applications. Its use of embedded C code, optimized for the processor, ensures high performance and efficiency, while the capacity to train on-device offers flexibility in various scenarios. The potential applications are vast, from smart sensors to real-time object detection, marking a significant leap in embedded AI technology.
Watch all my videos from the Embedded World 2024 here: https://www.youtube.com/playlist?list=PL7xXqJFxvYvjgUpdNMBkGzEWU6YVxR8Ga
Description by Chatgpt.
This video was filmed using the DJI Pocket 3 ($669 at https://amzn.to/4aMpKIC), watch all my DJI Pocket 3 videos here https://www.youtube.com/playlist?list=PL7xXqJFxvYvhDlWIAxm_pR9dp7ArSkhKK
Please support my work. You can donate any amount Value for Value! to my https://paypal.me/charbax or even better (no fees) to my wise account: charbax@gmail.com (you can join Wise for free at https://wise.com/invite/dic/nicolasc1941)
If you consider renting out a room or your home or your second home to make extra income, please use my link to sign up to Airbnb https://www.airbnb.com/r/ncharbonnier?s=6&t=061n0g I would get a €295 commission from Airbnb, when you list your space and complete a first qualifying stay. You will get €23, too. I appreciate this very much! I absolutely love Airbnb, it has changed the lives of millions of people around the world! Watch this video where I talk about Airbnbs in Turkey for example https://www.youtube.com/watch?v=g5rkFKThBew
Please Join as an Early Access Member on my YouTube channel at https://www.youtube.com/charbax/join and for about $5/month you get instant Early Access to all my videos immediately as I upload all my videos nightly from the conferences that I attend, instantly accessible to my members as I share them publicly at the rate of about 2 or 3 videos per day over the weeks following the show.")
