deepinsight/insightface · error · ValueError

no face or embedding

Error message

no face or embedding

What it means

Raised by _embedding_for_image when face selection returned None or the selected face has no normed_embedding. Note the surrounding try/except appends the error to the errors list and returns None, so this message typically surfaces in result error rows rather than as a crash.

Source

Thrown at python-package/insightface/gui/core/evaluation.py:228

def _embedding_for_image(
    path: Path,
    engine: FaceEngine,
    cache: Dict[str, np.ndarray],
    errors: List[Dict[str, Any]],
    stage: str,
    multi_face_policy: str = MULTI_FACE_REQUIRE_ONE,
) -> Optional[np.ndarray]:
    key = str(path)
    if key in cache:
        return cache[key]
    policy = _normalize_multi_face_policy(multi_face_policy)
    try:
        image, faces = _detect_faces_for_image(path, engine)
        face = _select_face_from_faces(faces, image.shape, path, policy, stage)
        if face is None or face.normed_embedding is None:
            raise ValueError("no face or embedding")
        embedding = np.asarray(face.normed_embedding, dtype=np.float32).reshape(-1)
        cache[key] = embedding
        return embedding
    except Exception as exc:
        errors.append({"path": key, "error": str(exc), "stage": stage})
        return None


def _collect_verification_specs(root: Path, auto_split: bool) -> tuple[List[Dict[str, Any]], List[Dict[str, Any]]]:
    identities = _identity_dirs(root)
    specs: List[Dict[str, Any]] = []
    structure_errors: List[Dict[str, Any]] = []
    if auto_split:
        for identity_dir in identities:
            identity = _identity_name(identity_dir)
            gallery, probes = _auto_split_identity_images(identity_dir)
            if gallery is None:
                structure_errors.append({"identity": identity, "error": "identity folder has no images"})

View on GitHub (pinned to 7fadd420c2)

Solutions

1. Check the errors rows (path/stage) to see which images failed and inspect them
2. Increase image resolution or face size (better quality source images)
3. Verify the FaceEngine model supports embedding extraction and was initialized correctly
4. Apply a different multi_face_policy or retry failing images individually

Example fix

# before
emb = _embedding_for_image(path, engine)
# after
emb = _embedding_for_image(path, engine)
if emb is None:
    print(result_errors_for(path))  # inspect {'path':..., 'error': 'no face or embedding', 'stage':...}

img = read_image(path)
faces = engine.detect_faces(img)
ok = len(faces) >= 1 and all(f.normed_embedding is not None for f in faces)

try:
    emb = _embedding_for_image(path, engine)
except ValueError as e:
    record_error(path, str(e))  # errors list already captures path/stage

When it happens

Trigger: An image where detection/select returns a face without normed_embedding, or selection yields None, during verification or identification evaluation runs.

Common situations: Very small or blurred faces where the model skips embedding extraction; engine/model misconfigured so embeddings are never computed; face crop too small after alignment.

Related errors

• skipped multi-face image ({face_count} faces)
• No gallery embeddings could be extracted.
• no face detected
• multiple faces detected ({face_count}); expected exactly one
• image read failure

AI-assisted analysis of deepinsight/insightface@7fadd420c2 (2026-08-28). Data as JSON: /api/errors/47d1964d2d86422a. Report an issue: GitHub.