roboflow/supervision · error · ValueError

top_k could not be calculated, confidence is None

Error message

top_k could not be calculated, confidence is None

What it means

Raised by `Classifications.get_top_k(k)` when `self.confidence` is None. Ranking by confidence is impossible without confidence scores, so the method refuses rather than returning arbitrary ordering. `Classifications` allows `confidence=None` at construction, but `get_top_k` does not.

Source

Thrown at src/supervision/classification/core.py:209

        Returns:
            A tuple containing the top k class IDs and confidences.

        Example:
            ```pycon
            >>> import numpy as np
            >>> import supervision as sv
            >>> classifications = sv.Classifications(
            ...     class_id=np.array([0, 1, 2]),
            ...     confidence=np.array([0.3, 0.9, 0.5])
            ... )
            >>> classifications.get_top_k(1)
            (array([1]), array([0.9]))

            ```
        """
        if self.confidence is None:
            raise ValueError("top_k could not be calculated, confidence is None")

        order = np.argsort(self.confidence)[::-1]
        top_k_order = order[:k]
        top_k_class_id = self.class_id[top_k_order]
        top_k_confidence = self.confidence[top_k_order]

        return top_k_class_id, top_k_confidence

View on GitHub (pinned to 7f254d9784)

Solutions

1. Supply confidence at construction: pass a probability array aligned with class_id.
2. If your model gives no scores, rank by another criterion yourself (e.g. class_id order) instead of calling `get_top_k`.
3. Guard the call: `if classifications.confidence is not None: ... get_top_k(k)`.

Example fix

# before
classifications = sv.Classifications(class_id=np.array([0, 1, 2]))
classifications.get_top_k(1)
# after
classifications = sv.Classifications(
    class_id=np.array([0, 1, 2]), confidence=np.array([0.3, 0.9, 0.5])
)
classifications.get_top_k(1)

if classifications.confidence is None:
    # rank by class_id order instead, or skip
    top = classifications.class_id[:k]
else:
    top_id, top_conf = classifications.get_top_k(k)

def has_confidence(c: sv.Classifications) -> bool:
    return c.confidence is not None

try:
    top_id, top_conf = classifications.get_top_k(k)
except ValueError:
    top_id = classifications.class_id[:k]; top_conf = None

When it happens

Trigger: Calling `sv.Classifications(class_id=np.array([0, 1, 2])).get_top_k(1)` — no confidence argument supplied at construction; calling `get_top_k` on classifications produced by a classifier connector that does not emit confidences.

Common situations: Using a label-only classifier output (e.g. CLIP zero-shot labels without probabilities) and then trying to rank; code paths shared between models where one model omits confidence.

Related errors

• confidence must be 1d np.ndarray with (n, ) shape
• No edges defined for class_id={class_id}.
• KeyPoints detection_confidence must be given for NMS to be e
• Detections confidence must be given for NMS to be executed.
• Detections confidence must be given for Soft-NMS to be execu

AI-assisted analysis of roboflow/supervision@7f254d9784 (2026-08-15). Data as JSON: /api/errors/6369dc9e57bd1862. Report an issue: GitHub.