non_max_suppression

non_max_suppression(boxlist: NpBoxList, max_output_size: int = 10000, iou_threshold: float = 1.0, score_threshold: float = -10.0) → NpBoxList

Non maximum suppression. This op greedily selects a subset of detection bounding boxes, pruning away boxes that have high IOU (intersection over union) overlap (> thresh) with already selected boxes. In each iteration, the detected bounding box with highest score in the available pool is selected.

Parameters:

• boxlist (BoxList) – BoxList holding N boxes. Must contain a ‘scores’ field representing detection scores. All scores belong to the same class.

• max_output_size (int) – Maximum number of retained boxes. Defaults to 10_000.

• iou_threshold (float) – Intersection over union threshold. Defaults to 1.0.

• score_threshold (float) – Minimum score threshold. Remove the boxes with scores less than this value. Default value is set to -10. A very low threshold to pass pretty much all the boxes, unless the user sets a different score threshold. Defaults to -10.0.

Returns:

A BoxList holding M boxes. where M <= max_output_size.

Return type:

BoxList

Raises:

• ValueError – If ‘scores’ field does not exist.

• ValueError – If threshold is not in [0, 1].

• ValueError – If max_output_size < 0.