Limited by 20minutes processing time

Question/Issue: My question is. Is there a way to upload the data, choose the training model but instead of training the model on edge impulse, trained it on my computer (locally)? Thanks

Project ID: 193597

Context/Use case:
Im building a vision assistance project for a robotic class and we need a large data set to be able to train properly the model. At the moment we have 6 classes (labels) but we need to add least 4 more and each class only has around 40-60images and we would like to add more. Because of this limitation, when the model is fully trained is no very accurate around 60%.

1 Like

Hi @victorinox

It sounds like you could benefit from using BYOM (bring your own model), you can train the model locally and then upload it to Edge Impulse to deploy any device.

You can also integrate Edge Impulse into your colab or local notebook with the Python SDK:




Hi Eoin,

Thank you very much for your answer :slight_smile: !