nncore.ops
Bbox
- nncore.ops.bbox.bbox_area(bboxes)[source]
Compute the areas of bounding boxes.
- Parameters:
bboxes (
nn.Tensor[N, 4]
) – Bounding boxes to be computed. They are expected to be in(x1, y1, x2, y2)
format.- Returns:
The computed areas.
- Return type:
nn.Tensor[N]
- nncore.ops.bbox.bbox_intersection(bboxes1, bboxes2, aligned=False)[source]
Compute the intersections among bounding boxes.
- Parameters:
bboxes1 (
nn.Tensor[N, 4]
) – Bounding boxes to be computed. They are expected to be in(x1, y1, x2, y2)
format.bboxes2 (
nn.Tensor[M, 4]
) – Bounding boxes to be computed. They are expected to be in(x1, y1, x2, y2)
format.aligned (bool, optional) – Whether to only compute the intersections among aligned bounding boxes. Default:
False
.
- Returns:
The computed intersection values.
- Return type:
nn.Tensor[N]
|nn.Tensor[N, M]
- nncore.ops.bbox.bbox_iou(bboxes1, bboxes2, aligned=False)[source]
Compute the intersection-over-unions (IoUs) among bounding boxes.
- Parameters:
bboxes1 (
nn.Tensor[N, 4]
) – Bounding boxes to be computed. They are expected to be in(x1, y1, x2, y2)
format.bboxes2 (
nn.Tensor[M, 4]
) – Bounding boxes to be computed. They are expected to be in(x1, y1, x2, y2)
format.aligned (bool, optional) – Whether to only compute the IoU among aligned bounding boxes. Default:
False
.
- Returns:
The computed pairwise IoU values.
- Return type:
nn.Tensor[N]
|nn.Tensor[N, M]
- nncore.ops.bbox.bbox_iof(bboxes1, bboxes2, aligned=False)[source]
Compute the intersection-over-foregrounds (IoFs) among bounding boxes.
- Parameters:
bboxes1 (
nn.Tensor[N, 4]
) – Bounding boxes to be computed. They are expected to be in(x1, y1, x2, y2)
format.bboxes2 (
nn.Tensor[M, 4]
) – Bounding boxes to be computed. They are expected to be in(x1, y1, x2, y2)
format.aligned (bool, optional) – Whether to only compute the IoF among aligned bounding boxes. Default:
False
.
- Returns:
The computed pairwise IoF values.
- Return type:
nn.Tensor[N]
|nn.Tensor[N, M]
- nncore.ops.bbox.remove_small_bboxes(bboxes, min_size)[source]
Remove bounding boxes which contains at least one side smaller than the minimum size.
- Parameters:
bboxes (
nn.Tensor[N, 4]
) – Bounding boxes to be computed. They are expected to be in(x1, y1, x2, y2)
format.min_size (float) – The minimum size of bounding boxes.
- Returns:
Indices of the bounding boxes that have both sides larger than
min_size
.- Return type:
nn.Tensor[K]
Temporal
- nncore.ops.temporal.temporal_area(windows)[source]
Compute the areas of temporal windows.
- Parameters:
windows (
nn.Tensor[N, 2]
) – Temporal windows to be computed. They are expected to be in(start, end)
format.- Returns:
The computed areas.
- Return type:
nn.Tensor[N]
- nncore.ops.temporal.temporal_intersection(windows1, windows2, aligned=False)[source]
Compute the intersections among temporal windows.
- Parameters:
windows1 (
nn.Tensor[N, 2]
) – Temporal windows to be computed. They are expected to be in(start, end)
format.windows2 (
nn.Tensor[M, 2]
) – Temporal windows to be computed. They are expected to be in(start, end)
format.aligned (bool, optional) – Whether to only compute the intersections among aligned temporal windows. Default:
False
.
- Returns:
The computed intersection values.
- Return type:
nn.Tensor[N]
|nn.Tensor[N, M]
- nncore.ops.temporal.temporal_iou(windows1, windows2, aligned=False)[source]
Compute the intersection-over-unions (IoUs) among temporal windows.
- Parameters:
windows1 (
nn.Tensor[N, 2]
) – Temporal windows to be computed. They are expected to be in(start, end)
format.windows2 (
nn.Tensor[M, 2]
) – Temporal windows to be computed. They are expected to be in(start, end)
format.aligned (bool, optional) – Whether to only compute the IoU among aligned temporal windows. Default:
False
.
- Returns:
The computed pairwise IoU values.
- Return type:
nn.Tensor[N]
|nn.Tensor[N, M]
- nncore.ops.temporal.temporal_iof(windows1, windows2, aligned=False)[source]
Compute the intersection-over-foregrounds (IoFs) among temporal windows.
- Parameters:
windows1 (
nn.Tensor[N, 2]
) – Temporal windows to be computed. They are expected to be in(start, end)
format.windows2 (
nn.Tensor[M, 2]
) – Temporal windows to be computed. They are expected to be in(start, end)
format.aligned (bool, optional) – Whether to only compute the IoF among aligned temporal windows. Default:
False
.
- Returns:
The computed pairwise IoF values.
- Return type:
nn.Tensor[N]
|nn.Tensor[N, M]