You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We routinely use CNNs for image classification. We should add a generic classifier to networks. We can use the layers.Encoder2D to build the model of arbitrary size/architecture, and then reduce to a dense layer with N outputs. The final layer should be the raw unscaled logits.
This would be a subclass of the Keras Model as described in the docs. In this way, we could build CNNs with arbitrary encoder networks, that reduce to a one-hot or binary classification output.
The text was updated successfully, but these errors were encountered:
We routinely use CNNs for image classification. We should add a generic classifier to
networks
. We can use thelayers.Encoder2D
to build the model of arbitrary size/architecture, and then reduce to a dense layer with N outputs. The final layer should be the raw unscaled logits.So we could build something like this:
This would be a subclass of the Keras
Model
as described in the docs. In this way, we could build CNNs with arbitrary encoder networks, that reduce to a one-hot or binary classification output.The text was updated successfully, but these errors were encountered: