Pack all blocks in a stage into a ``ResLayer``.

Dictionary mapper. Renames keys according to keymap provided. Args: d (dict): old dict keymap (dict): {'old_key':'new

Dictionary mapper. Renames keys according to keymap provided. Args: d (dict): old dict keymap (dict): {'old_k

Compute mask target for positive proposals in multiple images. Args: pos_proposals_list (list[Tensor]): Positive proposals in multiple

Merge hyperparameters into policy config. Only merge partial hyperparameters required of the policy. Args: policy (dict): Original p

Compute the moving average of the parameters using exponential moving average.

nn.Module: normalization layer after the second convolution layer

Apply Observation-Centric Momentum (OCM) to assign ids. OCM adds movement direction consistency into the association cost matrix. Thi

Change from point offset to point coordinate.

Pad gt_masks to shape of batch_input_shape.

Panoptic segmengation inference. Args: mask_cls (Tensor): Classfication outputs of shape (num_queries, cls_out_ch

Parse basic annotation for support and query set. Args: anno (dict): Annotation for single example. Return:

Parse basic annotation for support and query set. Args: anno (dict): Annotation for single example. coco (COCO): The

Parse raw annotation to target format. This method will return a dict which contains the data information of a sample. Args:

Split images into non-overlapped patches. Args: imgs (torch.Tensor): A batch of images, of shape B x C x H x W. Returns:

Permute and then flatten a tensor, from size (N, A, C, H, W) to (N, H * W * A, C). Args: layer (Tensor): Tensor of shape (N, C, H

Prepare the inputs of the Transformer. The forward procedure of the transformer is defined as: 'pre_transformer' -> 'encoder' -> 'pre

Extract elements necessary to represent a prediction into a dictionary. It's better to contain only basic data elements such as strin

Predict results by fusing the results of instance and semantic segmentations. Args: mask_results_list (list[:obj:`Instanc

Transform a batch of output features extracted from the head into bbox results. Note: When score_factors is not None, the cls_scores

Transform network output for a batch into bbox predictions. Args: center_heatmap_preds (list[Tensor]): Center predict heatmaps fo

Transform a batch of output features extracted by the head into bbox results. Args: cls_scores (list[Tensor]): Classificat

Transform network outputs for a batch into bbox predictions. Args: layer_cls_scores (Tensor): Classification outputs of the last

Perform forward propagation of the mask head and predict detection results on the features of the upstream network. Args:

Prepare data for the subsequent pipeline. Args: video_infos (dict): The whole video information. sampled_inds (list[i

Preprocessing the prototype before predict.

Used in meta testing. This function will be called before the meta testing. Obtain the vector based on the prototype. - torch.Tensor:

Preprocess the ground truth for all images. Args: batch_gt_instances (list[:obj:`InstanceData`]): Batch of gt_ins

Process one batch of data samples and predictions. The processed results should be stored in ``self.results``, which will be used to c

Process one batch of data samples and predictions. The processed results should be stored in ``self.results``, which will be used to c

Geometric transformat boxes in-place. Args: homography_matrix (Tensor or np.ndarray]): Shape (3, 3) for geometric

Geometric transformat boxes in-place. Args: homography_matrix (Tensor or np.ndarray]): Shape (3, 3) for geometric

The function to wrap the image and prompt. Make sure that len(prompt) == img_embeds.shape[0]. Args: img_embeds (torch.Te

Simple version of ``kwimage.Boxes.random`` Returns: Tensor: shape (n, 4) in x1, y1, x2, y2 format. References: https://gitla

Generate the mask for MAE Pre-training. Args: x (torch.Tensor): Image with data augmentation applied, which is of

Read a SN3 file in "Pascal Vincent" format (Lush file 'libidx/idx- io.lsh'). Argument may be a filename, compressed filename, or file object.

Predict bucketing estimation (cls_pred) and fine regression (offset pred) with side-aware features.

Register HuggingFace-style PreTrainedModel class.

See :func:`BaseInstanceMasks.resize`.

Get relative positional embeddings according to the relative positions of query and key sizes. Args: q_size (int): size of query q.

Resize relative position bias table. Args: src_shape (int): The resolution of downsampled origin training image, in format (H

Dump the detection results to a COCO style json file. There are 3 types of results: proposals, bbox predictions, mask predictions, an

Dump the detection results to a COCO style json file. There are 3 types of results: proposals, bbox predictions, mask predictions, an

The forward function that also takes the RFP features as input.

Scale RoI coordinates by scale factor. Args: rois (Tensor): RoI (Region of Interest), shape (n, 5) scale_factor (floa

Run NER on a caption and return the tokens and noun phrases. Args: caption (str): The input caption. Returns: Tuple[List, Lis

Sanitizes the input coordinates so that x1 < x2, x1 != x2, x1 >= 0, and x2 <= image_size. Also converts from relative to absolute coor

Saving predictions as json file for evaluation.

Performs the scaled all_reduce operation on the provided tensors. The input tensors are modified in-place. Currently supports only the sum re

Get updated `self_attn_mask` for distinct queries selection, it is used in self attention layers of decoder. Args: refere

Semantic segmengation postprocess. Args: mask_cls (Tensor): Classfication outputs of shape (num_queries, cls_out_

Set ``gt_score``.

Setup cache size limit of dynamo. Note: Due to the dynamic shape of the loss calculation and post-processing parts in the object detection al

Shear the BitmapMasks. Args: out_shape (tuple[int]): Shape for output mask, format (h, w). magnitude (int | float): T

Shift predictions to the original image. Args: det_data_samples (List[:obj:`DetDataSample`]): A list of patch results. offsets (S

Perform a single attention call between the visual and language inputs. Args: visual (Tensor): The visual input tensor.

Smooth L1 loss. Args: pred (Tensor): The prediction. target (Tensor): The learning target of the prediction. beta (float,

Split outputs of the denoising part and the matching part. For the total outputs of `num_queries_total` length, the former `num_denoi

Convert masks to the format of ndarray.

Transform function to perform photometric distortion on images. Args: results (dict): Result dict from loading pipeline.

Transform function to expand images, bounding boxes, masks, segmentation map. Args: results (dict): Result dict from load

Transform function to crop images and bounding boxes with minimum IoU constraint. Args: results (dict): Result dict from

Visualize image classification result. This method will draw an text box on the input image to visualize the information about image

Class names of PASCAL VOC.

Warp bounding boxes according to the warping matrix.

weighted boxes fusion <https://arxiv.org/abs/1910.13302> is a method for fusing predictions from different object detection models, which utilizes

Args: x: (B, H, W, C) Returns: windows: (num_windows*B, window_size, window_size, C)

Get KNN based on the pairwise distance. Args: x: (batch_size, num_dims, num_points, 1) y: (batch_size, num_dims, num_points, 1)

Convert ``xyxy`` style bounding boxes to ``xywh`` style for COCO evaluation. Args: bbox (numpy.ndarray): The bounding box

Convert ``xyxy`` style bounding boxes to ``xywh`` style for COCO evaluation. Args: bbox (numpy.ndarray): The bounding box

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

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