Change categories mapping logic (#3946)

* change pre-filtering logic

* Update src/otx/core/data/pre_filtering.py

Co-authored-by: Eunwoo Shin <eunwoo.shin@intel.com>

---------

Co-authored-by: Eunwoo Shin <eunwoo.shin@intel.com>

src/otx/core/data/pre_filtering.py

@@ -88,7 +88,15 @@ def remove_unused_labels(dataset: DmDataset, data_format: str, ignore_index: int
         raise ValueError(msg)
     if len(used_labels) == len(original_categories):
         return dataset
+    if data_format == "arrow" and max(used_labels) != len(original_categories) - 1:
+        # we assume that empty label is always the last one. If it is not explicitly added to the dataset,
+        # (not in the used labels) it will be filtered out.
+        mapping = {cat: cat for cat in original_categories[:-1]}
+    elif data_format == "arrow":
+        # this mean that some other class wasn't annotated, we don't need to filter the object classes
+        return dataset
+    else:
+        mapping = {original_categories[idx]: original_categories[idx] for idx in used_labels}
     msg = "There are unused labels in dataset, they will be filtered out before training."
     warnings.warn(msg, stacklevel=2)
-    mapping = {original_categories[idx]: original_categories[idx] for idx in used_labels}
     return dataset.transform("remap_labels", mapping=mapping, default="delete")