Make divisible function. This function rounds the channel number to the nearest value that can be divisible by the divisor. It is taken from

Merge augmented detection bboxes and scores. Args: aug_bboxes (list[Tensor]): shape (n, 4*#class) aug_scores (list[Tensor] or Non

nn.Module: normalization layer after the third convolution layer

Parse the package dependencies listed in a requirements file but strips specific versioning information. Args: fname (str): path to r

Refine bboxes during training. Args: rois (Tensor): Shape (n*bs, 5), where n is image number per GPU, and bs is t

Convert objects of various python types to :obj:`torch.Tensor`. Supported types are: :class:`numpy.ndarray`, :class:`torch.Tensor`, :class:`S

Box head forward function used in both training and testing.

Check whether the center of each box is in the patch. Args: patch (list[int]): The cropped area, [left, top, right, bottom].

Get final border for the target size. This function generates a ``final_border`` according to image's shape. The area between ``final

Mask head forward function used in both training and testing.

Get the flag indicate whether anchor centers are inside regions.

Map bboxes from testing scale to original image scale.

Build linear layer. Args: cfg (None or dict): The linear layer config, which should contain: - type (str): Layer type.

Calculate region of the box defined by the ratio, the ratio is from the center of the box to every edge.

Calculate a proportional bbox region. The bbox center are fixed and the new h' and w' is h * ratio and w * ratio. Args: bbox (Tensor

This function would modify some filed to keep the compatibility of config. For example, it will move some args which will be deprecated to th

Clip boxes dynamically for onnx. Since torch.clamp cannot have dynamic `min` and `max`, we scale the boxes by 1/max_shape and clamp in the

Encode bitmap mask to RLE code. Args: mask_results (list | tuple[list]): bitmap mask results. In mask scoring rcnn, mask_resu

Directly extract features from the backbone+neck.

Extract features from multiple images. Args: imgs (list[torch.Tensor]): A list of images. The images are augmente

Returns an available device, cpu, cuda or mlu.

Get user-defined property.

Get k of TopK for onnx exporting. The K of TopK in TensorRT should not be a Tensor, while in ONNX Runtime it could be a Tensor.Due to dynam

Load anns with the specified ids. self.anns is a list of annotation lists instead of a list of annotations. Args:

Load proposal from proposal file.

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

Test model with multiple gpus. This method tests model with multiple gpus and collects the results under two different modes: gpu and cpu mod

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.

nn.Module: the normalization layer named "norm2"

Converting the points set into bounding box. :param pts: the input points sets (fields), each points set (fields) is represented

Prepare an example input image for ``generate_inputs_and_wrap_model``. Args: input_config (dict): customized config describing the exampl

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

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

Make sure different ranks share the same seed. All workers must call this function, otherwise it will deadlock. This method is generally used

See :func:`BaseInstanceMasks.to_tensor`.

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

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

Box head forward function used in both training and testing. Returns all regression, classification results and a intermediate feature.

Box head forward function used in both training and testing.

Build head for each branch.

Crop image with a given center and size, then paste the cropped image to a blank image with two centers align. This function is equiv

Decode init_proposal_bboxes according to the size of images and expand dimension of init_proposal_features to batch_size. Args:

Paste instance masks according to boxes. This implementation is modified from https://github.com/facebookresearch/detectron2/ Args:

Initialize classification conv layers of the head.

Initialize bbox regression conv layers of the head.

Mask head forward function used in both training and testing.

Run forward function and calculate loss for mask head in training.

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

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

Assign w.r.t. the overlaps of bboxes with gts. Args: overlaps (Tensor): Overlaps between k gt_bboxes and n bboxes,

Test det bboxes with test time augmentation, can be applied in DenseHead except for ``RPNHead`` and its variants, e.g., ``GARPNHead``,

Augmented forward test function.

The key correspondence from bboxes to labels, masks and segmentations.

Build DistributedDataParallel module by device type. If device is cuda, return a MMDistributedDataParallel model; if device is mlu, return a

build DataParallel module by device type. if device is cuda, return a MMDataParallel model; if device is mlu, return a MLUDataParallel model.

Builder for Transformer.

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

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

Map from level to values.

Coerces input into a random number generator. If the input is None, then a global random state is returned. If the input is a numeric value,

Calculate recalls. Args: gts (list[ndarray]): a list of arrays of shape (n, 4) proposals (list[ndarray]): a list of arrays of sha

Evaluation in COCO protocol. Args: results (list[list | tuple]): Testing results of the dataset. metric (str | list[s

Instance segmentation and object detection evaluation in COCO protocol. Args: results (list[list | tuple | dict]): Testin

Find bboxes as long as a part of bboxes is inside the image. Args: bboxes (Tensor): Shape (N, 4). img_h (int): Image height.

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

Generate the coordinate. Args: featmap_sizes (tuple): The feature to be calculated, of shape (N, C, W, H). device (st

Get annotation by index. Args: idx (int): Index of data. Returns: dict: Annotation info of specified index.

Get category ids of concatenated dataset by index. Args: idx (int): Index of data. Returns: list[int]: All c

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

Get palette from various inputs. Args: palette (list[tuple] | str | tuple | :obj:`Color`): palette inputs. num_classes (int): the

Get multi-image mask results. Args: mlvl_mask_preds (list[Tensor]): Multi-level mask prediction. Each element in

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

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

Inverse function of sigmoid. Args: x (Tensor): The tensor to do the inverse. eps (float): EPS avoid numerical

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

Map from level to values based on max_value.

Load annotation from COCO style annotation file. Args: ann_file (str): Path of annotation file. Returns: lis

Log image size. Args: img_scale (tuple): Image size to be logged. shape_order (str, optional): The order of image shape.

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.