Train a TinyML suitable model

Origami-Cloudless-AI / TinyMLaaS-2023-winter

Run Hello World of TensorFlow Lite for micro automatically in Docker

https://Origami-TinyML.github.io/tflm_hello_world

Apache License 2.0

1 stars 2 forks source link

Train a TinyML suitable model #110

Open ArttuLe opened 1 year ago

ArttuLe commented 1 year ago

The current model doesn't fit into the memory of a microcontroller.

Research a TinyML compatible Neural network architecture (e.g., MobileNet v.X.x ) and train a new model that can be used on the devices.

doyu commented 1 year ago

Or if the model is a single purpose, you could explore its arch fully optimizaed only for your dataset. This could be generally smaller than making use of existing models.

ArttuLe commented 1 year ago

Modified the NN architecture a bit and now the resulting model should fit easily into the Arduino. @FexbYk23 can now use our own model in the device application.