RoiAlign - version 16#

This page documents version 16 of operator RoiAlign. See RoiAlign for the latest version (since version 22).

  • Domain: ai.onnx

  • Since version: 16

Region of Interest (RoI) align operation described in the Mask R-CNN paper. RoiAlign consumes an input tensor X and region of interests (rois) to apply pooling across each RoI; it produces a 4-D tensor of shape (num_rois, C, output_height, output_width).

RoiAlign is proposed to avoid the misalignment by removing quantizations while converting from original image into feature map and from feature map into RoI feature; in each ROI bin, the value of the sampled locations are computed directly through bilinear interpolation.

Inputs

  • X (T1): Input data tensor from the previous operator; 4-D feature map of shape (N, C, H, 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 (T1): RoIs (Regions of Interest) to pool over; rois is 2-D input of shape (num_rois, 4) given as [[x1, y1, x2, y2], …]. The RoIs’ coordinates are in the coordinate system of the input image. Each coordinate set has a 1:1 correspondence with the ‘batch_indices’ input.

  • batch_indices (T2): 1-D tensor of shape (num_rois,) with each element denoting the index of the corresponding image in the batch.

Outputs

  • Y (T1): RoI pooled output, 4-D tensor of shape (num_rois, C, output_height, output_width). The r-th batch element Y[r-1] is a pooled feature map corresponding to the r-th RoI X[r-1].

Type Constraints

  • T1: Constrain types to float tensors. Allowed types: tensor(bfloat16), tensor(double), tensor(float), tensor(float16).

  • T2: Constrain types to int tensors. Allowed types: tensor(int64).

Examples#

test_cc_roialign

Attributes:
  mode = "avg"
  output_height = 5
  output_width = 5
  sampling_ratio = 2
  spatial_scale = 1.0

Inputs:
  X: shape=(1, 1, 10, 10), dtype=float32
    [[[[0.  , 0.01, 0.02, ..., 0.07, 0.08, 0.09],
       [0.1 , 0.11, 0.12, ..., 0.17, 0.18, 0.19],
       [0.2 , 0.21, 0.22, ..., 0.27, 0.28, 0.29],
       ...,
       [0.7 , 0.71, 0.72, ..., 0.77, 0.78, 0.79],
       [0.8 , 0.81, 0.82, ..., 0.87, 0.88, 0.89],
       [0.9 , 0.91, 0.92, ..., 0.97, 0.98, 0.99]]]]
  rois: shape=(2, 4), dtype=float32
    [[0., 0., 9., 9.],
     [2., 2., 7., 7.]]
  batch_indices: shape=(2,), dtype=int64
    [0, 0]

Outputs:
  Y: shape=(2, 1, 5, 5), dtype=float32
    [[[[0.04674999, 0.0645    , 0.0825    , 0.10049999, 0.11849999],
       [0.22424999, 0.24199998, 0.26      , 0.278     , 0.296     ],
       [0.40425003, 0.422     , 0.44      , 0.458     , 0.47599998],
       [0.58425   , 0.60199994, 0.61999995, 0.63799995, 0.6559999 ],
       [0.7642499 , 0.78199995, 0.7999999 , 0.81799996, 0.8359999 ]]],


     [[[0.22      , 0.22999999, 0.24      , 0.25      , 0.26      ],
       [0.32      , 0.32999998, 0.33999997, 0.35000002, 0.36      ],
       [0.42      , 0.43      , 0.44      , 0.45      , 0.45999998],
       [0.52      , 0.53      , 0.53999996, 0.5500001 , 0.56      ],
       [0.62      , 0.63      , 0.64      , 0.65      , 0.65999997]]]]

test_cc_roialign_max

Attributes:
  mode = "max"
  coordinate_transformation_mode = "output_half_pixel"
  output_height = 5
  output_width = 5
  sampling_ratio = 2
  spatial_scale = 1.0

Inputs:
  X: shape=(1, 1, 10, 10), dtype=float32
    [[[[0.  , 0.01, 0.02, ..., 0.07, 0.08, 0.09],
       [0.1 , 0.11, 0.12, ..., 0.17, 0.18, 0.19],
       [0.2 , 0.21, 0.22, ..., 0.27, 0.28, 0.29],
       ...,
       [0.7 , 0.71, 0.72, ..., 0.77, 0.78, 0.79],
       [0.8 , 0.81, 0.82, ..., 0.87, 0.88, 0.89],
       [0.9 , 0.91, 0.92, ..., 0.97, 0.98, 0.99]]]]
  rois: shape=(2, 4), dtype=float32
    [[0., 0., 9., 9.],
     [2., 2., 7., 7.]]
  batch_indices: shape=(2,), dtype=int64
    [0, 0]

Outputs:
  Y: shape=(2, 1, 5, 5), dtype=float32
    [[[[0.14849998, 0.16649999, 0.18449998, 0.20249999, 0.22049998],
       [0.32849997, 0.34649998, 0.3645    , 0.3825    , 0.40049997],
       [0.5084999 , 0.5264999 , 0.5445    , 0.5625    , 0.5804999 ],
       [0.6884999 , 0.7065    , 0.72449994, 0.74249995, 0.76049995],
       [0.86849993, 0.88649994, 0.9044999 , 0.92249995, 0.9404999 ]]],


     [[[0.3025    , 0.3125    , 0.3225    , 0.3325    , 0.3425    ],
       [0.4025    , 0.4125    , 0.42249998, 0.4325    , 0.4425    ],
       [0.50249994, 0.5125    , 0.52250004, 0.5325    , 0.5425    ],
       [0.60249996, 0.61249995, 0.6225    , 0.6325    , 0.64250004],
       [0.7025    , 0.7125    , 0.72249997, 0.7325    , 0.74249995]]]]

Differences with previous version (10)#

SchemaDiff: RoiAlign (domain 'ai.onnx')

  • old version: 10

  • new version: 16

  • breaking: no