I wonder to know how edge impulse train the dataset, is it use increasment training on a pretrained model, like object detection, maybe use Yolo or other pretained model, and apply user labeled dataset to do add-on training and give quick response to end-user.
Hi @huntershuai,
It sounds like you are asking about “transfer learning.” If that is the case, then yes, we use transfer learning if you select the “Object Detection” block or “Transfer Learning (Images)” block for image classification. These start with a pre-trained model and update the parameters using a user-provided labeled dataset. Hope that helps!