EfficientNMS_TRT#
EfficientNMS_TRT - 1#
Version
domain: main
since_version: 1
function:
support_level:
shape inference:
This version of the operator has been available since version 1.
Summary
Efficient NMS TensorRT Plugin.
Attributes
background_class (required): Background class ID. Default value is
?.box_coding (required): Encoding type for the boxes or anchors inputs. Default value is
?.iou_threshold (required): Box IOU threshold value. Default value is
?.max_output_boxes (required): Max detections to output. Default value is
?.plugin_version (required): Version number of the TRT plugin. Default value is
?.score_activation (required): Activation function to apply to the scores input. Default value is
?.score_threshold (required): Score threshold value. Default value is
?.
Inputs
Between 2 and 3 inputs.
boxes (heterogeneous) - T: The boxes input tensor.
scores (heterogeneous) - T: The scores input tensor.
anchors (optional, heterogeneous) - T: The anchors input tensor.
Outputs
num_detections (heterogeneous) - tensor(int32): The num_detections output tensor.
detection_boxes (heterogeneous) - T: The detection_boxes output tensor.
detection_scores (heterogeneous) - T: The detection_scores output tensor.
detection_classes (heterogeneous) - tensor(int32): The detection_classes output tensor.
Examples