Compute losses of the head. Args: cls_scores (list[Tensor]): Box scores for each scale level Has shape (N, num_an

Compute loss of the head. Args: cls_scores (list[Tensor]): Box iou-aware scores for each scale level, each is a 4

Calculate the loss based on the features extracted by the detection head. Args: cls_scores (list[Tensor]): Box scores for

Make plugins for ResNet ``stage_idx`` th stage. Currently we support to insert ``context_block``, ``empirical_attention_block``, ``no

Run forward function and calculate loss for mask head in training. Args: x (tuple[Tensor]): Tuple of multi-level img features.

Convert [N, L, C] shape tensor to [N, C, H, W] shape tensor. Args: x (Tensor): The input tensor of shape [N, L, C] before conversion.

Converting the points set into bounding box. Args: pts (Tensor): the input points sets (fields), each points set

Test segment without test-time aumengtation. Only the output of last decoder layers was used. Args: mask_cls_results (Te

Predict results from a batch of inputs and data samples with post- processing. Args: batch_inputs (Tensor): Inputs, has s

Transform a batch of output features extracted from the head into bbox results. Args: rois (tuple[Tensor]): Tuple of boxe

Process one batch of data samples. The processed results should be stored in ``self.results``, which will be used to computed the met

Project state distribution to measurement space. Args: mean (ndarray): The state's mean vector (8 dimensional array).

Regress the bbox for the predicted class. Used in Cascade R-CNN. Args: priors (Tensor): Priors from `rpn_head` or last stage

Rename the key names in loss dict by adding a prefix. Args: prefix (str): The prefix for loss components. losses (dict): A dicti

Downsample the first feat and upsample last feat in feats. Args: x (tuple[Tensor]): Features from the upstream network, each is

Reweight losses in the dict by weight. Args: losses (dict): A dictionary of loss components. weight (float): Weight for loss com

Split side-aware features aligned with orders of bucketing targets.

Generate sparse points according to the ``prior_idxs``. Args: prior_idxs (Tensor): The index of corresponding anchors

unfold_wo_center, used in original implement in BoxInst: https://github.com/aim-uofa/AdelaiDet/blob/\ 4a3a1f7372c35b48ebf5f6adc59f135a0fa28d6

Update data root according to env MMDET_DATASETS. If set env MMDET_DATASETS, update cfg.data_root according to MMDET_DATASETS. Otherwise, usi

Args: img_data_sample (DetDataSample): single image output. step (int): The index of the current image.

Construct a layernorm module in the TF style (epsilon inside the square root).

Add shared or separable branch. convs -> avg pool (optional) -> fcs

Add shared or separable branch. convs -> avg pool (optional) -> fcs

build fc layers.

Box head forward function used in both training and testing. Args: x (tuple[Tensor]): List of multi-level img features.

Build head for each branch.

Check whether the `num_classes` in head matches the length of `CLASSES` in `dataset`. Args: runner (obj:`Runner`): runner

Expands attention_mask from ``[B, L_s]`` to ``[B, 1, L_t, L_s]``. Where ``B`` is batch_size, `L_s`` is the source sequence length, and ``L_t`

Set a fake roi head config.

Fuse the parameters in a branch with a conv and bn. Args: branch (mmcv.runner.Sequential): A branch with conv and bn. Re

Fuse the parameters in a branch with a conv and bn. Args: branch (mmcv.runner.Sequential): A branch with conv and bn. Re

Returns the index of the transform in a pipeline. If the transform is not found, returns -1.

Initialize classification conv layers of the head.

Join ``self.data_root`` with the specified path. If the path is an absolute path, just return the path. And if the path is None, retu

load data from CUB txt file, the every line of the file is idx and a data item.

Mask head forward function used in both training and testing. Args: stage (int): The current stage in Cascade RoI Head.

Mask head forward function used only in training. Args: stage (int): The current stage in Cascade RoI Head. x (tuple[

Mask head forward function used in both training and testing. Args: stage (int): The current stage in Cascade RoI Head.

Generate mesh grid of x and y. Args: x (torch.Tensor): Grids of x dimension. y (torch.Tensor): Grids of y dimension.

Generate mesh grid of x and y. Args: x (torch.Tensor): Grids of x dimension. y (torch.Tensor): Grids of y dimension.

calculate base classification task metrics, such as precision, recall, f1_score, support.

Transform a single image's features extracted from the head into bbox results. Args: cls_score_list (list[Tensor]): Box s

Record the homography matrix for the Resize.

Regularly record memory information. Args: runner (:obj:`Runner`): The runner of the training or evaluation proce

split cls_score. Args: cls_score (Tensor): The prediction with shape (N, C + 2). Returns: Tuple[Tensor, Tens

Run after every testing iterations. Args: runner (:obj:`Runner`): The runner of the testing process. batch_idx (int):

aligned bilinear, used in original implement in CondInst: https://github.com/aim-uofa/AdelaiDet/blob/\ c0b2092ce72442b0f40972f7c6dda8bb52c46d

Get anchor centers from anchors. Args: anchors (Tensor): Anchor list with shape (N, 4), ``xyxy`` format. Returns:

Convert annotation which can be polygons, uncompressed RLE to RLE. :return: binary mask (numpy 2D array)

Assign gt to priors. The assignment is done in following steps 1. assign -1 by default 2. compute the L1 cost between boxes.

Assign gt to points. This method assign a gt bbox to every points set, each points set will be assigned with the background_label (-

Assign gt to priors using SimOTA. Args: pred_instances (:obj:`InstanceData`): Instances of model predictions. It

Computes 1-to-k gt-pred matching based on the weighted costs. This method assign each query prediction to a ground truth or backgroun

Assign ids. Args: ids (list[int]): Tracking ids. det_bboxes (Tensor): of shape (N, 4) det_labels (Tensor)

Convert bbox coordinates from (cx, cy, ratio, h) to (x1, y1, x2, y2). Args: bbox (Tensor): Shape (n, 4) for bboxes. Returns:

Flip bboxes horizontally or vertically. Args: bboxes (Tensor): Shape (..., 4*k) img_shape (Tuple[int]): Image shape. dire

Synchronizing norm.

Convert masks from the form of bitmaps to polygons. Args: bitmap (ndarray): masks in bitmap representation. Return: list[nda

Classification-Aware Regression Loss (CARL). Args: cls_score (Tensor): Predicted classification scores. labels (Tensor): Targets

Calculate the centerness between anchors and gts. Only calculate pos centerness targets, otherwise there may be nan. Args:

Collected results in distributed environments. different from the function mmengine.dist.collect_results, tracking compute metrics don't use p

Async context manager that waits for work to complete on given CUDA streams.

Convert self to another box type. Args: dst_type (str or type): destination box type. Returns: :obj:`BaseBox

cearte a (conv + bn) Sequential layer.

See :func:`BaseInstanceMasks.crop`.

Expand ~ and ~user constructions. If user or $HOME is unknown, do nothing.