Remapping class is the operation of mapping original class labels in a dataset to other labels according to new class definition rules. It is commonly used to merge fine-grained classes, adapt to the class system of target models, and simplify multi-class tasks, serving as a key step in dataset standardization.
