ngraph.multiclass_nms

ngraph.multiclass_nms(boxes: Union[_pyngraph.Node, int, float, numpy.ndarray], scores: Union[_pyngraph.Node, int, float, numpy.ndarray], roisnum: Optional[Union[_pyngraph.Node, int, float, numpy.ndarray]] = None, sort_result_type: Optional[str] = 'none', sort_result_across_batch: Optional[bool] = False, output_type: Optional[str] = 'i64', iou_threshold: Optional[float] = 0.0, score_threshold: Optional[float] = 0.0, nms_top_k: Optional[int] = - 1, keep_top_k: Optional[int] = - 1, background_class: Optional[int] = - 1, nms_eta: Optional[float] = 1.0, normalized: Optional[bool] = True) → _pyngraph.Node

Return a node which performs MulticlassNms.

Parameters

  • boxes – Tensor with box coordinates.

  • scores – Tensor with box scores.

  • roisnum – Tensor with roisnum. Specifies the number of rois in each image. Required when ‘scores’ is a 2-dimensional tensor.

  • sort_result_type – Specifies order of output elements, possible values: ‘class’: sort selected boxes by class id (ascending) ‘score’: sort selected boxes by score (descending) ‘none’: do not guarantee the order.

  • sort_result_across_batch – Specifies whenever it is necessary to sort selected boxes across batches or not

  • output_type – Specifies the output tensor type, possible values: ‘i64’, ‘i32’

  • iou_threshold – Specifies intersection over union threshold

  • score_threshold – Specifies minimum score to consider box for the processing

  • nms_top_k – Specifies maximum number of boxes to be selected per class, -1 meaning to keep all boxes

  • keep_top_k – Specifies maximum number of boxes to be selected per batch element, -1 meaning to keep all boxes

  • background_class – Specifies the background class id, -1 meaning to keep all classes

  • nms_eta – Specifies eta parameter for adpative NMS, in close range [0, 1.0]

  • normalized – Specifies whether boxes are normalized or not

Returns

The new node which performs MuticlassNms