tensorflow/models · error · ValueError

The masks must have the same dimensions.

Error message

The masks must have the same dimensions.

What it means

Shape guard in calculate_iou: element-wise np.logical_and/or require mask1 and mask2 to have identical shapes. Mismatched shapes (masks from images resized differently, or a full-image mask compared against a cropped detection mask) make intersection/union undefined, so the function refuses before computing. Callers like find_similar_masks hit this when masks come from heterogeneous sources.

Solutions

1. Normalize all masks to the same shape (e.g. same image resolution) before comparison
2. In find_similar_masks, ensure candidate masks are generated from the same preprocessed image as the reference
3. Resize or pad the smaller mask to match the larger one's dimensions when semantically valid

When it happens

Trigger: Thrown at official/projects/waste_identification_ml/model_inference/postprocessing.py:295 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of tensorflow/models@e006f5f0d5 (2026-08-24). Data as JSON: /api/errors/e5a23b5ea5cd3bff. Report an issue: GitHub.

Appendix: source

Thrown at official/projects/waste_identification_ml/model_inference/postprocessing.py:295

  return score, box, combined_label


def calculate_iou(
    mask1: np.ndarray, mask2: np.ndarray
) -> Tuple[float, np.ndarray]:
  """Calculates the intersection over union (IoU) score for two masks.

  Args:
    mask1: The first mask.
    mask2: The second mask.

  Returns:
    The IoU scorea and union of two masks.
  """

  # Check if the masks have the same dimensions.
  if mask1.shape != mask2.shape:
    raise ValueError('The masks must have the same dimensions.')

  intersection = np.logical_and(mask1, mask2)
  union = np.logical_or(mask1, mask2)
  iou_score = np.sum(intersection) / np.sum(union)
  return iou_score, union


def find_similar_masks(
    results_1: DetectionResult,
    results_2: DetectionResult,
    num_detections: int,
    min_score_thresh: float,
    category_indices: List[List[Any]],
    category_index_combined: Dict[int, ItemDict],
    area_threshold: float,
    iou_threshold: float = 0.8,
) -> Dict[str, np.ndarray]:
  """Aligns the masks of the detections in `results_1` and `results_2`.

View on GitHub (pinned to e006f5f0d5)