palette does not exist, random is used by default. You can…

Error message

palette does not exist, random is used by default. You can also set the palette to customize.

What it means

Warning from _load_weights_to_model: after checking args.palette, config dataset metainfo, and checkpoint, no palette was found anywhere, so 'random' palette is assigned for visualization colors.

Solutions

Pass palette explicitly: DetInferencer(..., palette='coco')
Add 'palette' to your dataset's METAINFO in the config
Accept random colors if appearance doesn't matter

Example fix

// before
inferencer = DetInferencer(cfg, weights=w)  # random palette
// after
inferencer = DetInferencer(cfg, weights=w, palette='coco')

Defensive patterns

Strategy: fallback

Validate before calling

palette = palette if palette != 'none' else 'random'  # pre-decide instead of relying on silent default

Prevention

Set palette explicitly in configs for deterministic visuals
Add palette to custom dataset METAINFO

When it happens

Trigger: Running DetInferencer where palette='none' (default), the test dataset metainfo has no 'palette' key, and the checkpoint/config dataset_meta also lacks palette (e.g. non-COCO dataset without palette in METAINFO).

Common situations: Custom datasets whose METAINFO omits 'palette'; converted checkpoints; results are correct but instance colors vary run to run.

Related errors

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/c9910d09e09fd687. Report an issue: GitHub.

Appendix: source

Thrown at mmdet/apis/det_inferencer.py:154

                          'result is calculated by the randomly initialized '
                          'model!')
            warnings.warn('weights is None, use COCO classes by default.')
            model.dataset_meta = {'classes': get_classes('coco')}

        # Priority:  args.palette -> config -> checkpoint
        if self.palette != 'none':
            model.dataset_meta['palette'] = self.palette
        else:
            test_dataset_cfg = copy.deepcopy(cfg.test_dataloader.dataset)
            # lazy init. We only need the metainfo.
            test_dataset_cfg['lazy_init'] = True
            metainfo = DATASETS.build(test_dataset_cfg).metainfo
            cfg_palette = metainfo.get('palette', None)
            if cfg_palette is not None:
                model.dataset_meta['palette'] = cfg_palette
            else:
                if 'palette' not in model.dataset_meta:
                    warnings.warn(
                        'palette does not exist, random is used by default. '
                        'You can also set the palette to customize.')
                    model.dataset_meta['palette'] = 'random'

    def _init_pipeline(self, cfg: ConfigType) -> Compose:
        """Initialize the test pipeline."""
        pipeline_cfg = cfg.test_dataloader.dataset.pipeline

        # For inference, the key of ``img_id`` is not used.
        if 'meta_keys' in pipeline_cfg[-1]:
            pipeline_cfg[-1]['meta_keys'] = tuple(
                meta_key for meta_key in pipeline_cfg[-1]['meta_keys']
                if meta_key != 'img_id')

        load_img_idx = self._get_transform_idx(
            pipeline_cfg, ('LoadImageFromFile', LoadImageFromFile))
        if load_img_idx == -1:
            raise ValueError(

View on GitHub (pinned to cfd5d3a985)