NonMaxSuppression - 10 vs 11#
Next section compares an older to a newer version of the same operator after both definition are converted into markdown text. Green means an addition to the newer version, red means a deletion. Anything else is unchanged.
NonMaxSuppression10 → NonMaxSuppression11
RENAMED
|
@@ -1 +1 @@
|
|
|
1
1
|
Filter out boxes that have high intersection-over-union (IOU) overlap with previously selected boxes.
|
|
2
2
|
Bounding boxes with score less than score_threshold are removed. Bounding box format is indicated by attribute center_point_box.
|
|
3
3
|
Note that this algorithm is agnostic to where the origin is in the coordinate system and more generally is invariant to
|
|
4
4
|
orthogonal transformations and translations of the coordinate system; thus translating or reflections of the coordinate system
|
|
5
5
|
result in the same boxes being selected by the algorithm.
|
|
6
6
|
The selected_indices output is a set of integers indexing into the input collection of bounding boxes representing the selected boxes.
|
|
7
7
|
The bounding box coordinates corresponding to the selected indices can then be obtained using the Gather or GatherND operation.
|
|
8
8
|
**Attributes**
|
|
9
9
|
* **center_point_box**:
|
|
10
10
|
Integer indicate the format of the box data. The default is 0. 0 -
|
|
11
11
|
the box data is supplied as [y1, x1, y2, x2] where (y1, x1) and (y2,
|
|
12
12
|
x2) are the coordinates of any diagonal pair of box corners and the
|
|
13
13
|
coordinates can be provided as normalized (i.e., lying in the
|
|
14
14
|
interval [0, 1]) or absolute. Mostly used for TF models. 1 - the box
|
|
15
15
|
data is supplied as [x_center, y_center, width, height]. Mostly used
|
|
16
16
|
for Pytorch models.
|
|
17
17
|
**Inputs**
|
|
18
18
|
Between 2 and 5 inputs.
|
|
19
19
|
* **boxes** (heterogeneous) - **tensor(float)**:
|
|
20
20
|
An input tensor with shape [num_batches, spatial_dimension, 4]. The
|
|
21
21
|
single box data format is indicated by center_point_box.
|
|
22
22
|
* **scores** (heterogeneous) - **tensor(float)**:
|
|
23
23
|
An input tensor with shape [num_batches, num_classes,
|
|
24
24
|
spatial_dimension]
|
|
25
25
|
* **max_output_boxes_per_class** (optional, heterogeneous) - **tensor(int64)**:
|
|
26
26
|
Integer representing the maximum number of boxes to be selected per
|
|
27
27
|
batch per class. It is a scalar. Default to 0, which means no
|
|
28
28
|
output.
|
|
29
29
|
* **iou_threshold** (optional, heterogeneous) - **tensor(float)**:
|
|
30
30
|
Float representing the threshold for deciding whether boxes overlap
|
|
31
31
|
too much with respect to IOU. It is scalar. Value range [0, 1].
|
|
32
32
|
Default to 0.
|
|
33
33
|
* **score_threshold** (optional, heterogeneous) - **tensor(float)**:
|
|
34
34
|
Float representing the threshold for deciding when to remove boxes
|
|
35
35
|
based on score. It is a scalar.
|
|
36
36
|
**Outputs**
|
|
37
37
|
* **selected_indices** (heterogeneous) - **tensor(int64)**:
|
|
38
38
|
selected indices from the boxes tensor. [num_selected_indices, 3],
|
|
39
39
|
the selected index format is [batch_index, class_index, box_index].
|