Neuton TinyML

Neuton TinyML is an AI tool that allows users to automatically build extremely tiny models without coding and embed them into microcontrollers and sensors. It offers a no-code automated Tiny AutoML platform that can create compact and accurate models without additional compression. Neuton's unique features include selective connections, automatic structure growth, and a patented global optimization algorithm. It is free for developers and has been used in various applications such as touch-free interaction for smartwatches and hand activity recognition. Neuton's approach allows for the creation of models with minimal size and without loss of accuracy, making it suitable for running on the smallest IoT devices.

Key Features:

1. No-Code Automated Tiny ML Platform: Neuton is a no-code Tiny AutoML platform that allows users to build compact and accurate models without coding. It offers a highly automated and transparent pipeline, eliminating the need for manual actions.

2. Exceptionally Tiny Models without Loss of Accuracy: Neuton's neural framework enables the creation of models with an average size of less than 1 KB. Even 8-bit microcontrollers can run robust neural networks with up to 10x smaller models compared to other algorithms.

3. Unique Features of Neuton Network Framework: Neuton's network framework includes selective connections, a unique patented global optimization algorithm, automatic neuron-by-neuron structure growth, and constant cross-validation. These features help create models with minimal size and high accuracy.

Use Cases:

1. Touch-free Interaction for Smartwatches: Neuton's ready-made solution allows for various gestures to interact with smartwatches and other wearables. It offers high accuracy (97%) with an inference time of less than 1.5 ms and a total footprint of approximately 4 KB.

2. Logistics: On-device Package Tracking: Neuton's solution can recognize seven different events that may occur during the package delivery process. It achieves an accuracy of 98% with an inference time of less than 1 ms and a total footprint of around 4 KB.

3. Hand Activity Recognition on the STMicro ISPU: Neuton's solution can recognize daily human activities with only 3 KB in total footprint on the ultra-low power ISPU sensor. It achieves an accuracy of 98% with an inference time of less than 1 ms.

Neuton TinyML is a powerful AI tool that enables the automatic creation of extremely tiny models without coding. With its unique features and no-code approach, Neuton empowers developers to embed compact and accurate models into microcontrollers and sensors. Whether it's touch-free interaction for smartwatches or hand activity recognition, Neuton's capabilities make it a valuable solution for various practical applications. By offering models with minimal size and without loss of accuracy, Neuton is revolutionizing the field of edge computing and bringing AI to the smallest IoT devices.