One-hot-encoded level assignment for each spatial feature?
One-hot-encoded level assignment for each spatial feature?