tensorflow/models · error · ValueError

encoded_boxes.shape[-1] is {:d}, but must be 4.

Error message

encoded_boxes.shape[-1] is {:d}, but must be 4.

What it means

Error "encoded_boxes.shape[-1] is {:d}, but must be 4." thrown in tensorflow/models.

Source

Thrown at official/vision/ops/box_ops.py:408

    return encoded_boxes


def decode_boxes(encoded_boxes, anchors, weights=None):
  """Decodes boxes.

  Args:
    encoded_boxes: a tensor whose last dimension is 4 representing the
      coordinates of encoded boxes in dy, dx, dh, dw in order.
    anchors: a tensor whose shape is the same as, or `broadcastable` to `boxes`,
      representing the coordinates of anchors in ymin, xmin, ymax, xmax order.
    weights: None or a list of four float numbers used to scale coordinates.

  Returns:
    decoded_boxes: a tensor whose shape is the same as `boxes` representing the
      decoded box targets.
  """
  if encoded_boxes.shape[-1] != 4:
    raise ValueError(
        'encoded_boxes.shape[-1] is {:d}, but must be 4.'
        .format(encoded_boxes.shape[-1]))

  with tf.name_scope('decode_boxes'):
    encoded_boxes = tf.cast(encoded_boxes, dtype=anchors.dtype)
    dy, dx, dh, dw = tf.split(encoded_boxes, 4, -1)
    if weights:
      dy /= weights[0]
      dx /= weights[1]
      dh /= weights[2]
      dw /= weights[3]
    dh = tf.math.minimum(dh, BBOX_XFORM_CLIP)
    dw = tf.math.minimum(dw, BBOX_XFORM_CLIP)

    anchor_ymin, anchor_xmin, anchor_ymax, anchor_xmax = tf.split(
        anchors, 4, -1)
    anchor_h = anchor_ymax - anchor_ymin
    anchor_w = anchor_xmax - anchor_xmin

View on GitHub (pinned to e006f5f0d5)

Solutions

1. Pass encoded_boxes whose last dimension is 4.
2. Check the box encoding so each encoded box has exactly 4 values.

When it happens

Trigger: Thrown at official/vision/ops/box_ops.py:408 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/577fa761e658afe0. Report an issue: GitHub.