deepinsight/insightface · error · ValueError

Identity folder root not found: {root}

Error message

Identity folder root not found: {root}

What it means

_identity_dirs verifies the dataset root before scanning identity folders for verification/identification evaluation: if the path doesn't exist or isn't a directory, it raises ValueError immediately. This guards against silently evaluating an empty dataset (every downstream caller — _collect_verification_specs, _collect_identification_specs, the gallery/probe splitters — starts here).

Source

Thrown at python-package/insightface/gui/core/evaluation.py:108

    if len(negatives) < max(20, int(1 / max(target_far, 1e-9))):
        return "insufficient data"
    thresholds = sorted({float(row["similarity"]) for row in rows if row.get("similarity") is not None}, reverse=True)
    best_tar = 0.0
    for threshold in thresholds:
        far = sum(1 for row in negatives if row["similarity"] >= threshold) / max(1, len(negatives))
        if far <= target_far:
            tar = sum(1 for row in positives if row["similarity"] >= threshold) / max(1, len(positives))
            best_tar = max(best_tar, tar)
    return best_tar


def _identity_name(folder: Path) -> str:
    return folder.name


def _identity_dirs(root: Path) -> List[Path]:
    if not root.exists() or not root.is_dir():
        raise ValueError(f"Identity folder root not found: {root}")
    return sorted(path for path in root.iterdir() if path.is_dir())


def _auto_split_identity_images(identity_dir: Path) -> tuple[Path | None, List[Path]]:
    images = sorted(list_images(identity_dir, recursive=True))
    if not images:
        return None, []
    gallery_candidates = [path for path in images if "gallery" in path.stem.lower()]
    gallery = sorted(gallery_candidates or images)[0]
    probes = [path for path in images if path != gallery]
    return gallery, probes


def _normalize_multi_face_policy(policy: str) -> str:
    policy = (policy or MULTI_FACE_REQUIRE_ONE).strip()
    return policy if policy in MULTI_FACE_POLICIES else MULTI_FACE_REQUIRE_ONE

View on GitHub (pinned to 7fadd420c2)

Solutions

1. Check Path(root).is_dir() before invoking and print Path(root).resolve() to catch CWD-relative surprises.
2. Verify the dataset actually downloaded/mounted and pass an absolute path.
3. Fix the typo in the root argument (compare with the dataset's actual directory name).

Example fix

# before
run_identity_verification_evaluation(Path('bins/lfw'), ...)  # ValueError: Identity folder root not found: bins/lfw

# after
from pathlib import Path
root = Path('bins/lfw').expanduser().resolve()
assert root.is_dir(), f'missing dataset at {root}'
run_identity_verification_evaluation(root, ...)

from pathlib import Path
root = Path(root).expanduser().resolve()
if not root.is_dir():
    raise SystemExit(f'dataset root missing: {root}')

def is_dataset_root(p) -> bool:
    from pathlib import Path
    p = Path(p)
    return p.is_dir() and any(x.is_dir() for x in p.iterdir())

try:
    run_identity_verification_evaluation(root, ...)
except ValueError as e:
    if 'Identity folder root not found' in str(e):
        # fix/download path then retry
        ...
    raise

When it happens

Trigger: Calling run_identity_verification_evaluation or any _collect_* helper with a root Path that is misspelled, points to a file, or hasn't been downloaded/mounted yet.

Common situations: Wrong --root CLI argument; running evaluation before downloading the dataset; relative path resolved against a different CWD; symlink to an unmounted network share.

Related errors

• No verification pairs could be generated from the selected i
• Dataset root not found: {root}
• HERR_INVALID_PARAM
• HERR_INVALID_IMAGE_STREAM_PARAM
• HERR_INVALID_FACE_FEATURE

AI-assisted analysis of deepinsight/insightface@7fadd420c2 (2026-08-28). Data as JSON: /api/errors/1609957c36e01242. Report an issue: GitHub.