PaddlePaddle/PaddleOCR · error · TypeError

Unsupported result type: {type(result_obj)}

Error message

Unsupported result type: {type(result_obj)}

What it means

TypeError from _extract_items in ocr_reference_run.py when the OCR pipeline result is neither a dict nor an object supporting __getitem__. The reference script expects a PaddleOCR result that can be treated as a mapping containing rec_texts/rec_scores/rec_polys (or dt_polys); any other shape is rejected before field access.

Source

Thrown at deploy/ios_demo/scripts/ocr_reference_run.py:78

    if hasattr(obj, "tolist"):
        return obj.tolist()
    if isinstance(obj, (list, tuple)):
        return [_numpy_to_python(x) for x in obj]
    if isinstance(obj, dict):
        return {k: _numpy_to_python(v) for k, v in obj.items()}
    elif isinstance(obj, (str, int, float, bool)):
        return obj
    return str(obj)


def _extract_items(result_obj: Any) -> List[Dict[str, Any]]:
    """Build a list of {polygon, text, score} from an OCR pipeline result."""
    if isinstance(result_obj, dict):
        res = result_obj
    elif hasattr(result_obj, "__getitem__"):
        res = dict(result_obj)
    else:
        raise TypeError(f"Unsupported result type: {type(result_obj)}")
    return _extract_items_from_res_dict(res)


def _extract_items_from_res_dict(res: Dict[str, Any]) -> List[Dict[str, Any]]:
    texts = res.get("rec_texts") or []
    scores = res.get("rec_scores")
    polys = res.get("rec_polys") or res.get("dt_polys") or []
    if hasattr(scores, "tolist"):
        scores = scores.tolist()
    items: List[Dict[str, Any]] = []
    n = min(len(texts), len(polys))
    for i in range(n):
        poly = polys[i]
        if hasattr(poly, "tolist"):
            poly = poly.tolist()
        score = float(scores[i]) if scores is not None and i < len(scores) else None
        items.append(
            {

View on GitHub (pinned to 2661c7c0ef)

Solutions

1. Inspect `type(result)` and, if it is a list, take the first element / re-wrap: many versions return `[{'rec_texts': ...}]`.
2. Pin the PaddleOCR version the reference script was written against (check deploy/ios_demo docs/requirements).
3. Convert the result to a dict before calling: `_extract_items(dict(result[0]))` or use result[0].json['res'] style access on newer APIs.
4. Update _extract_items_from_res_dict if your version uses different keys (e.g. rec_polys vs dt_polys).

Example fix

// before
result = pipeline.predict(img)
items = _extract_items(result)  # TypeError on newer PaddleOCR

// after
result = pipeline.predict(img)
first = result[0] if isinstance(result, list) else result
res = first.get("res", first) if isinstance(first, dict) else dict(first)
items = _extract_items(res)

def is_extractable_result(result_obj) -> bool:
    return isinstance(result_obj, dict) or hasattr(result_obj, "__getitem__")

from typing import Any

def is_ocr_res_dict(obj: Any) -> bool:
    """True when obj can be treated as a mapping with OCR result keys."""
    if isinstance(obj, list) and len(obj) == 1:
        obj = obj[0]
    return isinstance(obj, dict) and any(
        k in obj for k in ("rec_texts", "rec_polys", "dt_polys")
    )

try:
    items = _extract_items(result)
except TypeError:
    # common PaddleOCR shape: list with a single result element
    first = result[0] if isinstance(result, list) else result
    items = _extract_items(dict(first.get("res", first)))

When it happens

Trigger: A PaddleOCR version whose predict() returns a bare list/tuple of result objects or a custom object without __getitem__; passing the wrong pipeline stage output (e.g. a detector-only result with dt_polys but consumed as a dict elsewhere); mocking the pipeline with a plain string.

Common situations: Upgrading PaddleOCR across major versions where the result API changed from list-of-lists to named dicts and back; switching between PP-OCRv3/v4 pipelines with different return conventions.

Related errors

• Detection model session is not initialized.
• Unexpected det output dims: [${dims.join(", ")}]
• Unexpected det output dims: [${od.join(", ")}]
• Detection batch output N=${String(nOut)} does not match inpu
• RecResizeImg.image_shape is required in rec inference.yml

AI-assisted analysis of PaddlePaddle/PaddleOCR@2661c7c0ef (2026-08-14). Data as JSON: /api/errors/4d171ff0527c5689. Report an issue: GitHub.