Uploaded October 2020 | Updated September 2026, 1 week ago
Not only can you run your ML model on a microcontroller, you can feed the data back into the model from the microcontroller, giving an ever improving distributed ML system...... More information and links below....
The idea of having the knowledge gained through machine learning on an embedded device is amazing in its own right.... but most of the things we want ML to train on comes from edge devices like the ESP32 and Nano33BLE / Teensy.
Being able to build this into a cyclical relationship, where the edge devices collect data, error check predictions, and feed the results back into storage.
The Build server then processes this using designed Keras model in Tensorflow, to produce the C++ data allowing the device to use the predictive model in its daily usage. It then also uploads the model back to the board(s), which can be scheduled, or on-demand depending on the project.
Being able to automate the data flow from multiple edge devices, to the central model, and re-deploy to devices in one workflow... makes for a powerful data collection and training of the AI, but also allows for constant comparison to, and learning from new data received by all nodes of the system.
Our Website:
visualmicro.com
Machine Learning Overview (All Examples linked from this page):
visualmicro.com/page/Machine-Learning-with-Arduino-and-TensorFlow.aspx?source=IWzch08NX7k
Arduino Cloud Build Instructable:
instructables.com/Visual-Micro-Cloud-Build
TensorFlow GitHub Repo:
github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/micro
TensorFlow Building Hello World:
github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/micro/examples/hello_world/train/README.md
#Arduino #VisualStudio #AI #MachineLearning
Not only can you run your ML model on a microcontroller, you can feed the data back into the model from the microcontroller, giving an ever improving distributed ML system...... More information and links below....
The idea of having the knowledge gained through machine learning on an embedded device is amazing in its own right.... but most of the things we want ML to train on comes from edge devices like the ESP32 and Nano33BLE / Teensy.
Being able to build this into a cyclical relationship, where the edge devices collect data, error check predictions, and feed the results back into storage.
The Build server then processes this using designed Keras model in Tensorflow, to produce the C++ data allowing the device to use the predictive model in its daily usage. It then also uploads the model back to the board(s), which can be scheduled, or on-demand depending on the project.
Being able to automate the data flow from multiple edge devices, to the central model, and re-deploy to devices in one workflow... makes for a powerful data collection and training of the AI, but also allows for constant comparison to, and learning from new data received by all nodes of the system.
Our Website:
visualmicro.com
Machine Learning Overview (All Examples linked from this page):
visualmicro.com/page/Machine-Learning-with-Arduino-and-TensorFlow.aspx?source=IWzch08NX7k
Arduino Cloud Build Instructable:
instructables.com/Visual-Micro-Cloud-Build
TensorFlow GitHub Repo:
github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/micro
TensorFlow Building Hello World:
github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/micro/examples/hello_world/train/README.md
#Arduino #VisualStudio #AI #MachineLearning
