cd.ops

This submodule contains PyTorch operations.

Drawing Operations

draw_contours_(canvas, contours, close=True)

Draw contours.

Draw contours on canvas.

Note

This is an inplace operation.

Parameters:

canvas – Tensor[h, w].
contours – Contours in (x, y) format. Tensor[num_contours, num_points, 2].
close – Whether to close contours. This is necessary if the last point of a contour is not equal to the first.

Box Operations

contours2boxes(contours, axis=-2)

Contours to boxes.

Converts contours to bounding boxes in (x0, y0, x1, y1) format.

Parameters:

contours – Contours as Tensor[(…, )num_points, 2]
axis – The num_points axis.

Returns:

nms(boxes, scores, thresh=0.5) → Tensor

Non-maximum suppression.

Perform non-maximum suppression (NMS) on the boxes according to their intersection-over-union (IoU).

Notes

Use torchvision.ops.boxes.nms if possible; This is just a “pure-python” alternative
cd.ops.boxes.nms for 8270 boxes: 13.9 ms ± 131 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)
tv.ops.boxes.nms for 8270 boxes: 1.84 ms ± 4.91 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)
cd.ops.boxes.nms for 179 boxes: 265 µs ± 1.4 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)
tv.ops.boxes.nms for 179 boxes: 103 µs ± 2.61 µs per loop (mean ± std. dev. of 7 runs, 10000 loops each)

Parameters:

boxes – Boxes. Tensor[num_boxes, 4] in (x0, y0, x1, y1) format.
scores – Scores. Tensor[num_boxes].
thresh – Threshold. Discards all overlapping boxes with IoU > thresh.

Returns:

Keep indices. Tensor[num_keep].

pairwise_box_iou(boxes1: Tensor, boxes2: Tensor) → Tensor

pairwise_generalized_box_iou(boxes1: Tensor, boxes2: Tensor) → Tensor

Losses

iou_loss(boxes, boxes_targets, reduction='mean', generalized=True, method='linear', min_size=None)

log_margin_loss(inputs: Tensor, targets: Tensor, m_pos=0.9, m_neg=None, exponent=1, reduction='mean', eps=1e-06)

margin_loss(inputs: Tensor, targets: Tensor, m_pos=0.9, m_neg=None, exponent=2, reduction='mean')

r1_regularization(logits, inputs, gamma=1.0, reduction='sum')

R1 regularization.

A gradient penalty regularization. This regularization may for example be applied to a discriminator with real data:

$R_1(\psi) &:= \frac{\gamma}{2} \mathbb E_{ p_{\mathcal D}(x)} \left[\|\nabla D_\psi(x)\|^2\right]$

References

Examples

>>> real.requires_grad_(True)
... real_logits = discriminator(real)
... loss_d_real = F.softplus(-real_logits)
... loss_d_r1 = r1_regularization(real_logits, real)
... loss_d_real = (loss_d_r1 + loss_d_real).mean()
... loss_d_real.backward()
... real.requires_grad_(False)

Parameters:

logits – Logits.
inputs – Inputs.
gamma – Gamma.
reduction – How to reduce all non-batch dimensions. E.g. 'sum' or 'mean'.

Returns:

Penalty Tensor[n].

reduce_loss(x: Tensor, reduction: str, **kwargs)

Reduce loss.

Reduces Tensor according to reduction.

Parameters:

x – Input.
reduction – Reduction method. Must be a symbol of torch.
**kwargs – Additional keyword arguments.

Returns:

Reduced Tensor.

Normalization

pixel_norm(x, dim=1, eps=1e-08)

Pixel normalization.

References

https://papers.nips.cc/paper/2012/file/c399862d3b9d6b76c8436e924a68c45b-Paper.pdf (“Local Repose Norm”)
https://arxiv.org/pdf/1710.10196.pdf

Parameters:

x – Input Tensor.
dim – Dimension to normalize.
eps – Epsilon.

Returns:

Normalized Tensor.

Common Operations

downsample_labels(inputs, size: List[int])

Down-sample via max-pooling and interpolation

Notes

Downsampling can lead to loss of labeled instances, both during max pooling and interpolation.
Typical timing: 0.08106 ms for 256x256

Parameters:

inputs – Label Tensor to resize. Shape (n, c, h, w)
size – Tuple containing target height and width.

Returns:

interpolate_vector(v, size, **kwargs)

Interpolate vector.

Parameters:

v – Vector as Tensor[d].
size – Target size.
**kwargs – Keyword arguments for F.interpolate

Returns:

minibatch_std_layer(x, channels=1, group_channels=None, epsilon=1e-08)

Minibatch standard deviation layer.

The minibatch standard deviation layer first splits the batch dimension into slices of size group_channels. The channel dimension is split into channels slices. For the groups the standard deviation is calculated and averaged over spatial dimensions and channel slice depth. The result is broadcasted to the spatial dimensions, repeated for the batch dimension and then concatenated to the channel dimension of x.

References

https://arxiv.org/pdf/1710.10196.pdf

Parameters:

x – Input Tensor[n, c, h, w].
channels – Number of averaged standard deviation channels.
group_channels – Number of channels per group. Default: batch size.
epsilon – Epsilon.

Returns:

Tensor[n, c + channels, h, w].

padded_stack2d(*images, dim=0) → Tensor

Padding stack.

Stacks 2d images along given axis. Spatial dimensions are padded according to largest height/width.

Parameters:

*images – Tensor[…, h, w]
dim – Stack dimension.

Returns:

Tensor

split_spatially(x, size)

Split spatially.

Splits spatial dimensions of Tensor x into patches of given size and adds the patches to the batch dimension.

Parameters:

x – Input Tensor[n, c, h, w, …].
size – Patch size of the splits.

Returns:

Tensor[n * h//height * w//width, c, height, width].

strided_upsampling2d(x, factor=2, const=0)

Strided upsampling.

Upsample by inserting rows and columns filled with constant.

Parameters:

x – Tensor[n, c, h, w].
factor – Upsampling factor.
const – Constant used to fill inserted rows and columns.

Returns:

Tensor[n, c, h*factor, w*factor].

CPN Operations

batched_box_nms(boxes: List[Tensor], scores: List[Tensor], *args, iou_threshold: float) → Tuple[List[Tensor], ...]

batched_box_nmsi(boxes: List[Tensor], scores: List[Tensor], iou_threshold: float) → List[Tensor]

fouriers2contours(fourier, locations, samples=64, sampling=None, cache: Dict[str, Tensor] | None = None, cache_size: int = 16)

Parameters:

fourier – Tensor[…, order, 4]
locations – Tensor[…, 2]
samples – Number of samples. Only used for default sampling, ignored otherwise.
sampling – Sampling t. Default is linspace 0..1. Device should match fourier and locations.
cache – Cache for initial zero tensors. When fourier shapes are consistent this can increase execution times.
cache_size – Cache size.

Returns:

Contours.

get_scale(actual_size, original_size, flip=True, dtype=torch.float32)

order_weighting(order, max_w=5, min_w=1, spread=None) → Tensor

refinement_bucket_weight(index, base_index)

rel_location2abs_location(locations, cache: Dict[str, Tensor] | None = None, cache_size: int = 16)

Parameters:

locations – Tensor[…, 2, h, w]. In xy format.
cache – can be None.
cache_size –

Returns:

remove_border_contours(contours, size, padding=1, top=True, right=True, bottom=True, left=True)

Remove border contours.

Remove contours that touch border regions.

Parameters:

contours – Contours as Tensor[num_contours, num_points, 2].
size – Context size.
padding – Padding. Determines the thickness of the border region. padding=1 removes all contours that overlap with the outer 1px frame.
top – Whether to test top border.
right – Whether to test right border.
bottom – Whether to test bottom border.
left – Whether to test left border.

Returns:

Keep mask as Tensor[num_contours].

resolve_refinement_buckets(samplings, num_buckets)

scale_contours(actual_size, original_size, contours)

Parameters:

actual_size – Image size. E.g. (256, 256)
original_size – Original image size. E.g. (512, 512)
contours – Contours that are to be scaled to from actual_size to original_size. E.g. array of shape (1, num_points, 2) for a single contour or tuple/list of (num_points, 2) arrays. Last dimension is interpreted as (x, y).

Returns:

Rescaled contours.

scale_fourier(actual_size, original_size, fourier, location)

Parameters:

actual_size – Image size. E.g. (256, 256)
original_size – Original image size. E.g. (512, 512)
fourier – Fourier descriptor. E.g. array of shape (…, order, 4).
location – Location. E.g. array of shape (…, 2). Last dimension is interpreted as (x, y).

Returns:

Rescaled fourier, rescaled location