.. _op_ai_onnx_MaxRoiPool:

MaxRoiPool
==========

- **Domain**: ``ai.onnx``
- **Since version**: 22

ROI max pool consumes an input tensor X and regions of interest (RoIs) to
apply max pooling across each RoI, to produce output 4-D tensor of shape
(num_rois, channels, pooled_shape[0], pooled_shape[1]).

**Inputs**

- **X** (*T*): Input data tensor from the previous operator; dimensions for image case are (N x C x H x W), where N is the batch size, C is the number of channels, and H and W are the height and the width of the data.
- **rois** (*T*): RoIs (Regions of Interest) to pool over. Should be a 2-D tensor of shape (num_rois, 5) given as [[batch_id, x1, y1, x2, y2], ...].

**Outputs**

- **Y** (*T*): RoI pooled output 4-D tensor of shape (num_rois, channels, pooled_shape[0], pooled_shape[1]).

**Attributes**

- **pooled_shape** (*int[]*): ROI pool output shape (height, width).
- **spatial_scale** (*float*): Multiplicative spatial scale factor to translate ROI coordinates from their input scale to the scale used when pooling.

**Type Constraints**

- **T**: Constrain input and output types to float tensors.
  Allowed types: tensor(bfloat16), tensor(double), tensor(float), tensor(float16).

Examples
--------

**test_cc_maxroipool_default**

.. code-block:: text

    Node:
      MaxRoiPool(X, rois) -> (Y)
      Attributes:
        pooled_shape = [2, 2]

.. code-block:: text

    Inputs:
      X: shape=(1, 2, 6, 6), dtype=float32
        [[[[0.        , 0.02777778, 0.05555556, 0.08333334, 0.11111111, 0.1388889 ],
           [0.16666667, 0.19444445, 0.22222222, 0.25      , 0.2777778 , 0.30555555],
           [0.33333334, 0.3611111 , 0.3888889 , 0.41666666, 0.44444445, 0.4722222 ],
           [0.5       , 0.5277778 , 0.5555556 , 0.5833333 , 0.6111111 , 0.6388889 ],
           [0.6666667 , 0.6944444 , 0.7222222 , 0.75      , 0.7777778 , 0.8055556 ],
           [0.8333333 , 0.8611111 , 0.8888889 , 0.9166667 , 0.9444444 , 0.9722222 ]],
        
          [[1.        , 1.0277778 , 1.0555556 , 1.0833334 , 1.1111112 , 1.1388888 ],
           [1.1666666 , 1.1944444 , 1.2222222 , 1.25      , 1.2777778 , 1.3055556 ],
           [1.3333334 , 1.3611112 , 1.3888888 , 1.4166666 , 1.4444444 , 1.4722222 ],
           [1.5       , 1.5277778 , 1.5555556 , 1.5833334 , 1.6111112 , 1.6388888 ],
           [1.6666666 , 1.6944444 , 1.7222222 , 1.75      , 1.7777778 , 1.8055556 ],
           [1.8333334 , 1.8611112 , 1.8888888 , 1.9166666 , 1.9444444 , 1.9722222 ]]]]
      rois: shape=(2, 5), dtype=float32
        [[0., 0., 0., 5., 5.],
         [0., 1., 1., 4., 4.]]

    Outputs:
      Y: shape=(2, 2, 2, 2), dtype=float32
        [[[[0.3888889 , 0.4722222 ],
           [0.8888889 , 0.9722222 ]],
        
          [[1.3888888 , 1.4722222 ],
           [1.8888888 , 1.9722222 ]]],
        
        
         [[[0.3888889 , 0.44444445],
           [0.7222222 , 0.7777778 ]],
        
          [[1.3888888 , 1.4444444 ],
           [1.7222222 , 1.7777778 ]]]]

**test_cc_maxroipool_spatial_scale**

.. code-block:: text

    Node:
      MaxRoiPool(X, rois) -> (Y)
      Attributes:
        pooled_shape = [3, 3]
        spatial_scale = 0.5

.. code-block:: text

    Inputs:
      X: shape=(1, 2, 6, 6), dtype=float32
        [[[[0.        , 0.02777778, 0.05555556, 0.08333334, 0.11111111, 0.1388889 ],
           [0.16666667, 0.19444445, 0.22222222, 0.25      , 0.2777778 , 0.30555555],
           [0.33333334, 0.3611111 , 0.3888889 , 0.41666666, 0.44444445, 0.4722222 ],
           [0.5       , 0.5277778 , 0.5555556 , 0.5833333 , 0.6111111 , 0.6388889 ],
           [0.6666667 , 0.6944444 , 0.7222222 , 0.75      , 0.7777778 , 0.8055556 ],
           [0.8333333 , 0.8611111 , 0.8888889 , 0.9166667 , 0.9444444 , 0.9722222 ]],
        
          [[1.        , 1.0277778 , 1.0555556 , 1.0833334 , 1.1111112 , 1.1388888 ],
           [1.1666666 , 1.1944444 , 1.2222222 , 1.25      , 1.2777778 , 1.3055556 ],
           [1.3333334 , 1.3611112 , 1.3888888 , 1.4166666 , 1.4444444 , 1.4722222 ],
           [1.5       , 1.5277778 , 1.5555556 , 1.5833334 , 1.6111112 , 1.6388888 ],
           [1.6666666 , 1.6944444 , 1.7222222 , 1.75      , 1.7777778 , 1.8055556 ],
           [1.8333334 , 1.8611112 , 1.8888888 , 1.9166666 , 1.9444444 , 1.9722222 ]]]]
      rois: shape=(1, 5), dtype=float32
        [[ 0.,  0.,  0., 10., 10.]]

    Outputs:
      Y: shape=(1, 2, 3, 3), dtype=float32
        [[[[0.19444445, 0.25      , 0.30555555],
           [0.5277778 , 0.5833333 , 0.6388889 ],
           [0.8611111 , 0.9166667 , 0.9722222 ]],
        
          [[1.1944444 , 1.25      , 1.3055556 ],
           [1.5277778 , 1.5833334 , 1.6388888 ],
           [1.8611112 , 1.9166666 , 1.9722222 ]]]]

Differences with previous version (1)
-------------------------------------

**SchemaDiff**: ``MaxRoiPool`` (domain ``'ai.onnx'``)

* old version: 1
* new version: 22
* breaking: no

**Type constraints:**

* changed 'T': added types: ['tensor(bfloat16)']

Version History
---------------

- :doc:`Version 1 <MaxRoiPool-1>`