Box filtering
powerboxes.remove_small_boxes(boxes, min_size)
Remove boxes with area less than min_area.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
boxes |
NDArray[T]
|
2d array of boxes in xyxy format |
required |
min_size |
float
|
minimum area of boxes to keep |
required |
Raises:
| Type | Description |
|---|---|
TypeError
|
if boxes is not numpy array |
Returns:
| Type | Description |
|---|---|
NDArray[T]
|
np.ndarray: 2d array of boxes in xyxy format |
powerboxes.nms(boxes, scores, iou_threshold, score_threshold)
Apply non-maximum suppression to boxes.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
boxes |
NDArray[T]
|
2d array of boxes in xyxy format |
required |
scores |
NDArray[float64]
|
1d array of scores |
required |
iou_threshold |
float
|
threshold for iou |
required |
score_threshold |
float
|
threshold for scores |
required |
Raises:
| Type | Description |
|---|---|
TypeError
|
if boxes or scores are not numpy arrays |
Returns:
| Type | Description |
|---|---|
NDArray[uint64]
|
npt.NDArray[np.uint64]: 1d array of indices to keep |
powerboxes.rtree_nms(boxes, scores, iou_threshold, score_threshold)
Apply non-maximum suppression to boxes.
Uses an rtree to speed up computation. This is only available for signed integer dtypes and float32 and float64.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
boxes |
NDArray[Union[float64, float32, int64, int32, int16]]
|
2d array of boxes in xyxy format |
required |
scores |
NDArray[float64]
|
1d array of scores |
required |
iou_threshold |
float
|
threshold for iou |
required |
score_threshold |
float
|
threshold for scores |
required |
Raises:
| Type | Description |
|---|---|
TypeError
|
if boxes or scores are not numpy arrays |
Returns:
| Type | Description |
|---|---|
NDArray[uint64]
|
npt.NDArray[np.uint64]: 1d array of indices to keep |