Register a box type. Args: name (str): The name of box type. box_type (type): Box mode class to be registered. force (boo

Register a box converter. Args: src_type (str or type): source box type name or class. dst_type (str or type): destination box ty

Resize bounding boxes with ``results['scale_factor']``.

Resize images with ``results['scale']``.

Resize masks with ``results['scale']``

Save the gt and prediction results.

Test using the metric in the same way as Evaluator.

Set type of tood head.

This function is used to transpose image first tensors to level first ones.

Provide additional adaptive positive samples to the classification branch. Args: flatten_points (Tensor): The point after

Add ground truth as assigned results. Args: gt_labels (torch.Tensor): Labels of gt boxes

Regularly record memory information. Args: runner (:obj:`Runner`): The runner of the testing process. batch_idx (int)

Regularly record memory information. Args: runner (:obj:`Runner`): The runner of the training process. batch_idx (int

Run after every ``self.interval`` validation iteration. Args: runner (:obj:`Runner`): The runner of the validation process.

Regularly record memory information. Args: runner (:obj:`Runner`): The runner of the validation process. batch_idx (i

Import a module from albumentations. It inherits some of :func:`build_from_cfg` logic. Args: cfg (dict): Config dict. It

Check if the parameters of the module is all zeros. Args: module (nn.Module): The module to be checked. Returns: bool: Wheth

Assign gt to bboxes. The assignment is done in following steps 1. compute alignment metric between all bbox (bbox of all pyramid

Attach a text label above each bar in *rects*, displaying its height.

Reversible Backward pass. Any intermediate activations from `buffer_layers` are recovered from ctx. Each layer implements its own loi

Activation re-compute with the following equation. x2 = y2 - g(y1), g = FFN x1 = y1 - f(x2), f = MSHA

Convert a concated label tensor to onehot format. Args: batch_label (torch.Tensor): A concated label tensor from multiple sam

Batch shuffle, for making use of BatchNorm. Args: x (torch.Tensor): Data in each GPU. Returns: Tuple[torch.Tensor, torch.Ten

Compute deltas of proposals w.r.t. gt. We usually compute the deltas of x, y, w, h of proposals w.r.t ground truth bboxes to get regression t

Perform forward propagation and loss calculation of the bbox head on the features of the upstream network. Args: x (tuple

Perform forward propagation and loss calculation of the bbox head on the features of the upstream network. Args: stage (i

Perform forward propagation and loss calculation of the bbox head on the features of the upstream network. Args: x (tuple

Map bboxes from the original image scale to testing scale.

Split batch bbox prediction back to each image.

Rescale bounding box w.r.t. scale_factor. Args: bboxes (Tensor): Shape (n, 4) for bboxes or (n, 5) for rois scale_factor (float):

Compute reweight_factor for regression and classification loss.

Calculate the average recall. Args: pred (torch.Tensor | np.ndarray | Sequence): The prediction results. A :obj:`

Calculate the precision, recall, f1-score and support. Args: pred (torch.Tensor | np.ndarray | Sequence): The prediction

Concatenate a sequence of masks into one single mask instance. Args: masks (Sequence[BitmapMasks]): A sequence of mask instances.

Concat a batch of label tensor to one tensor. Args: elements (List[tensor]): A batch of labels. Returns: Tuple[torch.Tensor,

Calculate the centroid coordinates of the mask. Args: mask (Tensor): The mask to be calculated, shape (h, w). esp (float): Avoid

Channel Shuffle operation. This function enables cross-group information flow for multiple groups convolution layers. Args: x (T

Perform text cleaning before tokenizer.

Compute the metrics from processed results. Args: results (list): The processed results of each batch. Returns:

Compute the metrics from processed results. Args: results (list): The processed results of each batch. Returns:

Compute the recall of occluded or separated masks. Args: result_dict (dict): Processed mask results. gt_ann (list): O

Set attend previous.

Set condition media locations.

Set condition vision features.

Construct the reconstruction target. In addition to splitting images into tokens, this module will also normalize the image according

Converts a sequence of [tokens|ids] using the vocab.

Convert CrowdHuman dataset in COCO style. Args: ann_dir (str): The path of CrowdHuman dataset. save_dir (str): The path to save a

Convert logits to class scores.

Convert box coordinates from corners ((x1, y1), (x2, y1), (x1, y2), (x2, y2)) to (x1, y1, x2, y2). Args: corners (Tensor)

Create a independent figure. Different from the :func:`plt.figure`, the figure from this function won't be managed by matplotlib. And it has

construct a map such that positive_map[i,j] = True if box i is associated to token j Args: tokenized: The tokenized input. to

Create a dictionary mapping the label to the token. Args: positive_map (Tensor): The positive map tensor. plus (int, optional): Va

Crop predicted masks by zeroing out everything not in the predicted bbox. Args: mask_preds (list[Tensor]): Predicted prot

Generate bbox and apply lambda correction. Args: img_shape (tuple): Image shape as tuple lam (float): Cutmix lambda v

Convert box coordinates from (cx, cy, w, h) to (x1, y1, x2, y2). Args: boxes (Tensor): cxcywh boxes tensor with shape of (..., 4)

Create sample results that can be passed to BBoxHead.get_targets.

Split and dispatch a model across devices. The function depends on the `accelerate` package. Refers to https://huggingface.co/docs/accelerate

Download object at the given URL to a local path. Modified from https://pytorch.org/docs/stable/hub.html#torch.hub.download_url_to_file

Update moving average.

Calculate the emd loss. Note: This implementation is modified from https://github.com/Purkialo/ CrowdDet/blob/master/

Get box regression transformation deltas that can be used to transform the ``bboxes`` into the ``gt_bboxes``. Args: bboxe

The function to encode the images.

Evaluate mAP of a dataset. Args: det_results (list[list]): [[cls1_det, cls2_det, ...], ...]. The outer list indicates images,

Evaluate the model performance of the whole dataset after processing all batches. Args: size (int): Length of the entire

Extract feature from the input `x`, and split the output to a list. Args: x (Tensor): of shape [N, C, H, W]. N is the number of p

Extract features from the input tensor with shape (N, C, ...). Args: inputs (Tensor): A batch of inputs. The shape of it should b

Evaluate proposal recall with COCO's fast_eval_recall. Args: results (List[dict]): Results of the dataset. proposal_n

Find inside box points. Boxes dimension must be 2. Args: points (Tensor): Points coordinates. Has shape of (m, 2). is

Convert various python types to score-format tensor. Supported types are: :class:`numpy.ndarray`, :class:`torch.Tensor`, :class:`Sequence`.

Forward features from the upstream network. Args: x (tuple[Tensor]): Features from the upstream network, each is

Forward feature from the upstream network to get prototypes and linearly combine the prototypes, using masks coefficients, into instan

Forward features from the upstream network. Args: x (tuple[Tensor]): Features from the upstream network, each is

Forward features from the upstream network. Args: x (Tuple[Tensor]): Features from the upstream network, each is

resize a batch of images and bboxes.

Sparse layer norm forward function with 2D data. Args: x (torch.Tensor): The input tensor. data_format (str): The for

Generate features for masked images. The function supports two kind of forward behaviors. If the ``mask`` is ``True``, the function w

Forward process for `ConditionalAttention`. Args: query (Tensor): The input query with shape [bs, num_queries, em

Forward with Transformer decoder. The forward procedure of the transformer is defined as: 'pre_transformer' -> 'encoder' -> 'pre_deco