open-mmlab/mmdetection · warning

The unexpected output indicates an issue with named entity r

Error message

The unexpected output indicates an issue with named entity recognition. You can try setting custom_entities=True and running again to see if it helps.

What it means

Grounding DINO predict found predicted label indices >= len(entity): the NER-parsed entity list doesn't cover all class slots, so a prediction can't be mapped to a name and is labeled 'unobject'.

Source

Thrown at mmdet/models/detectors/grounding_dino.py:610

                data_samples.token_positive_map = token_positive_maps[i]

            head_inputs_dict = self.forward_transformer(
                visual_feats, text_dict, batch_data_samples)
            results_list = self.bbox_head.predict(
                **head_inputs_dict,
                rescale=rescale,
                batch_data_samples=batch_data_samples)

        for data_sample, pred_instances, entity, is_rec_task in zip(
                batch_data_samples, results_list, entities, is_rec_tasks):
            if len(pred_instances) > 0:
                label_names = []
                for labels in pred_instances.labels:
                    if is_rec_task:
                        label_names.append(entity)
                        continue
                    if labels >= len(entity):
                        warnings.warn(
                            'The unexpected output indicates an issue with '
                            'named entity recognition. You can try '
                            'setting custom_entities=True and running '
                            'again to see if it helps.')
                        label_names.append('unobject')
                    else:
                        label_names.append(entity[labels])
                # for visualization
                pred_instances.label_names = label_names
            data_sample.pred_instances = pred_instances
        return batch_data_samples

View on GitHub (pinned to cfd5d3a985)

Solutions

1. Use custom_entities=True with an explicit entity list
2. Rewrite the caption as period-separated phrases

Example fix

# before
results = detector(inputs, texts=['red car and blue bus on street'])
# after
results = detector(inputs, texts=['red car . blue bus .'], custom_entities=True)

entities = [e for e in my_categories]  # explicit list
caption = ' . '.join(my_categories)
results = detector(inputs, texts=[caption], custom_entities=True)

When it happens

Trigger: Predicting with a caption where entity parsing yields fewer entities than the positive map implies (re-caption task skips this; recognition task hits it).

Common situations: Free-form captions with unusual punctuation; captions not following the 'phrase . phrase .' convention.

Related errors

• The unexpected output indicates an issue with named entity r
• If you want to reduce GPU memory usage,
• There is not post_norm in {name}
• Inputting a text that is too long will result in poor predic
• The annotation file of Open Images Challenge should be a txt

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/3926cba7c25f16d7. Report an issue: GitHub.