have a pre-trained network that can object detect one class. How do I add more classes to this model without losing the previous knowledge of that original class?
I have read in another post that suggests creating a new model, with the "base model" layers except for the last two layers. (Why the last 2 specifically?). and setting the new model last 2 layers weights to the base model's last 2 layers weights. Link to Post: How to add a new category to a deep learning model
weights_training = base_model.layers[-2].get_weights()
new_model.layers[-2].set_weights(weights_training)
note that the above code is in tensorflow not tfjs.
Would I be able to do the same in tensorflow.js?