Extract features. Args: batch_inputs (Tensor): Image tensor, has shape (bs, dim, H, W). Returns: tuple[Tenso

Extract features. Args: batch_inputs (Tensor): Image tensor with shape (N, C, H ,W). Returns: tuple[Tensor]:

Extract features. Args: batch_inputs (Tensor): Image tensor with shape (N, C, H ,W). Returns: tuple[Tensor]:

Extract features. Args: batch_inputs (Tensor): Image tensor with shape (N, C, H ,W). Returns: tuple[Tensor]:

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

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

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

Extract features from the input dict. Args: images (tensor, optional): The images to extract features. Defaults t

The function to extract image latent features.

The function to extract image latent features.

Extract roi features. Args: feats (list[Tensor]): list of multi-level image features. bboxes (list[Tensor]): list of

Extract roi features. Args: feats (list[Tensor]): list of multi-level image features. bboxes (list[Tensor]): list of

The function to extract text latent features.

The function to extract text latent features.

Flip boxes horizontally or vertically in-place. Args: img_shape (Tuple[int, int]): A tuple of image height and width.

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

Forward function. Args: inputs (tuple[Tensor]): Features from the upstream network, each is a 4D-tensor.

Forward process of Transformer, which includes four steps: 'pre_transformer' -> 'encoder' -> 'pre_decoder' -> 'decoder'. We summarized

fuse conv and bn.

Gather feature according to index. Args: feat (Tensor): Target feature map. ind (Tensor): Target coord index. mask (Tenso

Generate 2D gaussian heatmap. Args: heatmap (Tensor): Input heatmap, the gaussian kernel will cover on it and maintain the ma

Get vid ids that satisfy given filter conditions. :param vidIds (int array) : get vids for given ids :param catIds (int array) : get

Get custom classification channels. Args: num_classes (int): The number of classes. Returns: int: The custom

Makes broadcastable attention and causal masks so that future and masked tokens are ignored. Arguments: attention_mask (:

Get label assignment (from teacher). Args: cls_scores (list[Tensor]): Box scores for each scale level Has shape (

Only keep loading image and annotations related configuration. Args: pipeline (list[dict]): Data pipeline configs. Returns:

Extract local maximum pixel with given kernel. Args: heat (Tensor): Target heatmap. kernel (int): Kernel size of max pooling. Def

Get the position embedding of proposal.

Get the reference points used in decoder. Args: spatial_shapes (Tensor): The shape of all feature maps, has shape

Compute regression and classification targets in multiple images. Args: gt_bboxes (list[Tensor]): Ground truth bboxes for each im

A wrapper for computing ATSS and FCOS targets for points in multiple images. Args: cls_scores (list[Tensor]): Box iou-awa

Get targets for ATSS head. This method is almost the same as `AnchorHead.get_targets()`. Besides returning the targets as the parent

Get the tokens positive and prompts for the caption.

Get the tokens positive and prompts for the caption.

Get top k positions from heatmap. Args: scores (Tensor): Target heatmap with shape [batch, num_classes, height, width].

Args: box_cls (Tensor): tensor of shape (batch_size, num_proposals, K). The tensor predicts the classification pr

Importance-based Sample Reweighting (ISR_P), positive part. Args: cls_score (Tensor): Predicted classification scores. bbox_pred

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

Loss function. Args: all_layers_cls_scores (Tensor): Classification scores of all decoder layers, has shape (num_

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

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

Calculate the loss based on the features extracted by the detection head. Args: tl_heats (list[Tensor]): Top-left corner

Calculate losses. Args: x_split (Tensor): The embed features belonging to key image. ref_x_split (Tensor): The embed

Calculate the loss of a single scale level based on the features extracted by the detection head. Args: anchors (Tensor):

Obtain tight bounding boxes of binary masks. Args: masks (Tensor): Binary mask of shape (n, h, w). Returns: Tensor: Bboxe wi

Matrix NMS for multi-class masks. Args: masks (Tensor): Has shape (num_instances, h, w) labels (Tensor): Labels of corresponding

A Wrapper of MSE loss. Args: pred (Tensor): The prediction. target (Tensor): The learning target of the prediction. Returns:

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

nn.Module: normalization layer after the first convolution layer

Process the predictions and visualization results from ``forward`` and ``visualize``. This method should be responsible for the follo

The process before the final classification head. The input ``feats`` is a tuple of tensor, and each tensor is the feature of a backb

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

predict_by_feat() is implemented in StageCascadeRPNHead.

Transform a batch of output features extracted from the head into mask results. Args: mask_preds (tuple[Tensor]): Tuple o

Process the inputs into a model-feedable format. Customize your preprocess by overriding this method. Preprocess should return an ite

transfer tensors in predictions to CPU.

Refine bboxes through stages.

Register a box type. A record will be added to ``bbox_types``, whose key is the box type name and value is the box type itself. Simultaneousl

Register a box converter. A record will be added to ``box_converter``, whose key is '{src_type_name}2{dst_type_name}' and value is the conver

Register HuggingFace-style PreTrainedTokenizerBase class.

See :func:`BaseInstanceMasks.rescale`.

Expand an array of boxes by a given scale. Args: bboxes (Tensor): Shape (m, 4) scale (float): The scale factor of bboxes Ret

Fix seeds to allow for stochastic elements such as dropout to be reproduced exactly in activation recomputation in the backward pass.

Set label of ``gt_label``.

Overrides this method to set the :attr:`device` Args: device (int or torch.device, optional): The desired device of the

Feedforward layers. Args: mult (int): Layer expansion muliplier. Defaults to 4.