tensorflow/models · error · ValueError

translate_y_only_bboxes does not support rank 4 boxes

Error message

translate_y_only_bboxes does not support rank 4 boxes

What it means

Error "translate_y_only_bboxes does not support rank 4 boxes" thrown in tensorflow/models.

Source

Thrown at official/vision/ops/augment.py:1642

  else:
    image = translate_y(image, pixels, replace)

  # Convert bbox coordinates to pixel values.
  image_height = tf.shape(image)[0]
  image_width = tf.shape(image)[1]
  # pylint:disable=g-long-lambda
  wrapped_shift_bbox = lambda bbox: _shift_bbox(
      bbox, image_height, image_width, pixels, shift_horizontal)
  # pylint:enable=g-long-lambda
  bboxes = tf.map_fn(wrapped_shift_bbox, bboxes)
  return image, bboxes


def translate_y_only_bboxes(
    image: tf.Tensor, bboxes: tf.Tensor, prob: float, pixels: int, replace):
  """Apply translate_y to each bbox in the image with probability prob."""
  if bboxes.shape.rank == 4:
    raise ValueError('translate_y_only_bboxes does not support rank 4 boxes')

  func_changes_bbox = False
  prob = _scale_bbox_only_op_probability(prob)
  return _apply_multi_bbox_augmentation_wrapper(
      image, bboxes, prob, translate_y, func_changes_bbox, pixels, replace)


def _randomly_negate_tensor(tensor):
  """With 50% prob turn the tensor negative."""
  should_flip = tf.cast(tf.floor(tf.random.uniform([]) + 0.5), tf.bool)
  final_tensor = tf.cond(should_flip, lambda: tensor, lambda: -tensor)
  return final_tensor


def _rotate_level_to_arg(level: float):
  level = (level / _MAX_LEVEL) * 30.
  level = _randomly_negate_tensor(level)
  return (level,)

View on GitHub (pinned to e006f5f0d5)

Solutions

1. Pass boxes as rank 3 or lower; rank-4 batched boxes are not supported here.
2. Reshape boxes to [num_boxes, 4] before calling this augmentation.

When it happens

Trigger: Thrown at official/vision/ops/augment.py:1642 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/7aa1833bf2c666b7. Report an issue: GitHub.