open-mmlab/mmdetection · error · ValueError

Unsupported input type: {type(single_input)}

Error message

Unsupported input type: {type(single_input)}

What it means

In DetInferencer.visualize, each input must be either a str path (or directory) or a numpy ndarray image. Any other type (e.g., a PIL Image, torch tensor, list, or bytes) falls into the else branch and raises 'Unsupported input type'. The type check happens per-element of the inputs list after inference, during the visualization loop.

Source

Thrown at mmdet/apis/det_inferencer.py:485

        if self.visualizer is None:
            raise ValueError('Visualization needs the "visualizer" term'
                             'defined in the config, but got None.')

        results = []

        for single_input, pred in zip(inputs, preds):
            if isinstance(single_input, str):
                img_bytes = mmengine.fileio.get(single_input)
                img = mmcv.imfrombytes(img_bytes)
                img = img[:, :, ::-1]
                img_name = osp.basename(single_input)
            elif isinstance(single_input, np.ndarray):
                img = single_input.copy()
                img_num = str(self.num_visualized_imgs).zfill(8)
                img_name = f'{img_num}.jpg'
            else:
                raise ValueError('Unsupported input type: '
                                 f'{type(single_input)}')

            out_file = osp.join(img_out_dir, 'vis',
                                img_name) if img_out_dir != '' else None

            self.visualizer.add_datasample(
                img_name,
                img,
                pred,
                show=show,
                wait_time=wait_time,
                draw_gt=False,
                draw_pred=draw_pred,
                pred_score_thr=pred_score_thr,
                out_file=out_file,
            )
            results.append(self.visualizer.get_image())
            self.num_visualized_imgs += 1

View on GitHub (pinned to cfd5d3a985)

Solutions

  1. Convert PIL images to numpy before calling: img = np.asarray(pil_img)[:, :, ::-1] (BGR) or pass RGB consistently with the pipeline's image loading convention
  2. Pass file paths (str) or directory paths instead of decoded objects
  3. Ensure every element of the inputs list is uniformly str or np.ndarray, not a mix

Example fix

// before
from PIL import Image
img = Image.open('a.jpg')
inferencer(img, show=True)  # ValueError
// after
import numpy as np
img = np.asarray(Image.open('a.jpg').convert('RGB'))
inferencer(img, show=True)
Defensive patterns

Strategy: type-guard

Validate before calling

import numpy as np
inputs = [np.asarray(x) if hasattr(x, 'convert') or hasattr(x, 'mode') else x for x in inputs]  # PIL -> np
assert all(isinstance(i, (str, np.ndarray)) for i in inputs), 'inputs must be str paths or np.ndarray'

Type guard

import numpy as np
def is_inferencer_input_ok(x) -> bool:
    return isinstance(x, (str, np.ndarray))

Try / catch

try:
    inferencer(inputs, show=True)
except ValueError as e:
    if 'Unsupported input type' in str(e):
        raise TypeError('Convert PIL/tensor inputs to np.ndarray or pass file paths') 
    raise

Prevention

When it happens

Trigger: Calling the inferencer with show/return_vis/img_out_dir enabled and inputs containing PIL.Image objects, torch tensors, raw file bytes, or nested lists instead of plain str paths or np.ndarray images.

Common situations: Wrapping the inferencer in a service that decodes images with PIL first, passing cv2.imread results mixed with PIL objects, or passing a numpy array of dtype object. Note visualize is only reached when visualization is requested; without it, __call__ preprocess may handle some types differently.

Related errors

AI-assisted analysis of open-mmlab/mmdetection@cfd5d3a985 (2026-08-27). Data as JSON: /api/errors/960c638af6e05298. Report an issue: GitHub.