This class represents masks in the form of bitmaps. Args: masks (ndarray): ndarray of masks in shape (N, H, W), where N is th

This class represents masks in the form of polygons. Polygons is a list of three levels. The first level of the list corresponds to objects,

ResNet backbone. Args: depth (int): Depth of resnet, from {18, 34, 50, 101, 152}. stem_channels (int | None): Number of stem chan

r"""Feature Pyramid Network. This is an implementation of paper `Feature Pyramid Networks for Object Detection <https://arxiv.org/abs/1612.03

Stores assignments between predicted and truth boxes. Attributes: num_gts (int): the number of truth boxes considered when computing this

A general data structure of OpenMMlab. A data structure that stores the meta information, the annotations of the images or the model predicti

r""" LayerNorm that supports two data formats: channels_last (default) or channels_first. The ordering of the dimensions in the inputs. channels_

Assign a corresponding gt bbox or background to each bbox. Each proposals will be assigned with `-1`, or a semi-positive integer indicating t

Standard points generator for multi-level (Mlvl) feature maps in 2D points-based detectors. Args: strides (list[int] | list[tuple[int

MobileNetV2 backbone. Args: widen_factor (float): Width multiplier, multiply number of channels in each layer by this amount.

BottleBlock for TridentResNet. Args: trident_dilations (tuple[int, int, int]): Dilations of different trident branch.

Inverted Residual Block. Args: in_channels (int): The input channels of this Module. out_channels (int): The output channels of t

ResLayer to build ResNet style backbone for RPF in detectoRS. The difference between this module and base class is that we pass ``rfp_inplane

Applies padding to input (if needed) so that input can get fully covered by filter you specified. It support two modes "same" and "corner". The

CSP-Darknet backbone used in YOLOv5 and YOLOX. Args: arch (str): Architecture of CSP-Darknet, from {P5, P6}. Default: P5.

Compose multiple transforms sequentially. Args: transforms (Sequence[dict | callable]): Sequence of transform object or confi

Image to Patch Embedding. We use a conv layer to implement PatchEmbed. Args: in_channels (int): The num of input channels. Default:

Swin Transformer A PyTorch implement of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows` - https://arxiv.

Handle empty batch dimension to AdaptiveAvgPool2d.

ResLayer to build ResNet style backbone. Args: block (nn.Module): block used to build ResLayer. inplanes (int): inplanes of block

Computes one-to-one matching between predictions and ground truth for mask. This class computes an assignment between the targets and the pre

Open Images dataset for detection. Args: ann_file (str): Annotation file path. label_file (str): File path of the label descripti

Merge patch feature map. This layer groups feature map by kernel_size, and applies norm and linear layers to the grouped feature map. Our imp

Try to convert inputs to FP16 and CPU if got a PyTorch's CUDA Out of Memory error. It will do the following steps: 1. First retry after c

Data structure for instance-level annnotations or predictions. Subclass of :class:`GeneralData`. All value in `data_fields` should have the s

Pyramid Vision Transformer (PVT) Implementation of `Pyramid Vision Transformer: A Versatile Backbone for Dense Prediction without Convolution

Extra layers of SSD backbone to generate multi-scale feature maps. Args: in_channels (Sequence[int]): Number of input channels per scale.

Simplified version of original basic residual block. This is used in `SCNet <https://arxiv.org/abs/2012.10150>`_. - Norm layer is now optiona

A wrapper of concatenated dataset. Same as :obj:`torch.utils.data.dataset.ConcatDataset`, but concat the group flag for image aspect ratio.

ResNet backbone for DetectoRS. Args: sac (dict, optional): Dictionary to construct SAC (Switchable Atrous Convolution). Defau

Randomly drop some regions of feature maps. Please refer to the method proposed in `DropBlock <https://arxiv.org/abs/1810.12890>`_ for deta

Dynamic ReLU (DyReLU) module. See `Dynamic ReLU <https://arxiv.org/abs/2003.10027>`_ for details. Current implementation is specialized for t

Extract RoI features from all level feature maps levels. This is the implementation of `A novel Region of Interest Extraction Layer for Insta

HourglassNet backbone. Stacked Hourglass Networks for Human Pose Estimation. More details can be found in the `paper <https://arxiv.org/a

Load an image from file. Required keys are "img_prefix" and "img_info" (a dict that must contain the key "filename"). Added or updated keys a

A wrapper of multiple images mixed dataset. Suitable for training on multiple images mixed data augmentation like mosaic and mixup. For the a

Implementation of `PVTv2: Improved Baselines with Pyramid Vision Transformer <https://arxiv.org/pdf/2106.13797.pdf>`_.

Squeeze-and-Excitation Module. Args: channels (int): The input (and output) channels of the SE layer. ratio (int): Squeeze ratio

The neck of YOLOV3. It can be treated as a simplified version of FPN. It will take the result from Darknet backbone and do some upsampling an

Standard anchor generator for 2D anchor-based detectors. Args: strides (list[int] | list[tuple[int, int]]): Strides of anchors

Assign a corresponding gt bbox or background to each bbox. Each proposals will be assigned with an integer indicating the ground truth index

This class is almost the same as official pycocotools package. It implements some snake case function aliases. So that the COCO class has the

The neck used in `CenterNet <https://arxiv.org/abs/1904.07850>`_ for object classification and box regression. Args: in_channel (int

Assign pixels at the center region of a bbox as positive. Each proposals will be assigned with `-1`, `0`, or a positive integer indicating th

Custom dataset for detection. The annotation format is shown as follows. The `ann` field is optional for testing. .. code-block:: none

Implements the decoder in DETR transformer. Args: return_intermediate (bool): Whether to return intermediate outputs. post_norm_c

FPG. Implementation of `Feature Pyramid Grids (FPG) <https://arxiv.org/abs/2004.03580>`_. This implementation only gives the basic struct

FPN_CARAFE is a more flexible implementation of FPN. It allows more choice for upsample methods during the top-down pathway. It can reproduce

HRNet backbone. `High-Resolution Representations for Labeling Pixels and Regions arXiv: <https://arxiv.org/abs/1904.04514>`_. Args:

FPN structure in NASFPN. Implementation of paper `NAS-FCOS: Fast Neural Architecture Search for Object Detection <https://arxiv.org/abs/1906.

NAS-FPN. Implementation of `NAS-FPN: Learning Scalable Feature Pyramid Architecture for Object Detection <https://arxiv.org/abs/1904.07392>`_

Random sampler. Args: num (int): Number of samples pos_fraction (float): Fraction of positive samples neg_pos_ub (int, op

Simplest RoI head, with only two fc layers for classification and regression respectively.

Cross Stage Partial Layer. Args: in_channels (int): The input channels of the CSP layer. out_channels (int): The output channels

r"""Channel Mapper to reduce/increase channels of backbone features. This is used to reduce/increase channels of backbone features. Args:

A wrapper of repeated dataset with repeat factor. Suitable for training on class imbalanced datasets like LVIS. Following the sampling strate

Coco dataset for Panoptic segmentation. The annotation format is shown as follows. The `ann` field is optional for testing. .. code-bloc

ModulatedDeformConv2d with normalization layer used in DyHead. This module cannot be configured with `conv_cfg=dict(type='DCNv2')` because Dy

Feature Adaption Module. Feature Adaption Module is implemented based on DCN v1. It uses anchor shape prediction rather than feature map to

High-Resolution Module for HRNet. In this module, every branch has 4 BasicBlocks/Bottlenecks. Fusion/Exchange is in this module.

Computes one-to-one matching between predictions and ground truth. This class computes an assignment between the targets and the predictions

Label Assignment Head from the paper: `Improving Object Detection by Label Assignment Distillation <https://arxiv.org/pdf/2108.10520.pdf>`_

Memory profiler hook recording memory information including virtual memory, swap memory, and the memory of the current process. Args:

r"""Online Hard Example Mining Sampler described in `Training Region-based Object Detectors with Online Hard Example Mining <https://arxiv.org

Wrapper for detector's inference with ONNXRuntime.

Open Images Challenge dataset for detection.

Image to Patch Embedding.

Assign a corresponding gt bbox or background to each point. Each proposals will be assigned with `0`, or a positive integer indicating the gr

RegNet backbone. More details can be found in `paper <https://arxiv.org/abs/2003.13678>`_ . Args: arch (dict): The parameter of RegN

A wrapper of repeated dataset. The length of repeated dataset will be `times` larger than the original dataset. This is useful when the data

r"""ResNetV1d variant described in `Bag of Tricks <https://arxiv.org/pdf/1812.01187.pdf>`_. Compared with default ResNet(ResNetV1b), ResNetV1

Bbox sampling result. Example: >>> # xdoctest: +IGNORE_WANT >>> from mmdet.core.bbox.samplers.sampling_result import * # NOQA

Position encoding with sine and cosine functions. See `End-to-End Object Detection with Transformers <https://arxiv.org/pdf/2005.12872>`_ for

Args: embed_dims (int): The feature dimension. num_heads (int): Parallel attention heads. feedforward_channels (int): The

The stem layer, stage 1 and stage 2 in Trident ResNet are identical to ResNet, while in stage 3, Trident BottleBlock is utilized to replace the

Uniform Matching between the anchors and gt boxes, which can achieve balance in positive anchors, and gt_bboxes_ignore was not considered for

AutoAssignHead head used in AutoAssign. More details can be found in the `paper <https://arxiv.org/abs/2007.03496>`_ . Args: for

Bidirectional Corner Pooling Module (TopLeft, BottomRight, etc.) Args: in_channels (int): Input channels of module. out_channels

Objects as Points Head. CenterHead use center_point to indicate object's position. Paper link <https://arxiv.org/abs/1904.07850> Args:

ConvUpsample performs 2x upsampling after Conv. There are several `ConvModule` layers. In the first few layers, upsampling will be applied af

Head of CornerNet: Detecting Objects as Paired Keypoints. Code is modified from the `official github repo <https://github.com/princeton-vl/Co

Decoupled SOLO mask head used in `SOLO: Segmenting Objects by Locations. <https://arxiv.org/abs/1912.04488>`_ Args: init_cfg (dict o

Decoupled Light SOLO mask head used in `SOLO: Segmenting Objects by Locations <https://arxiv.org/abs/1912.04488>`_ Args: with_dcn (bo

Delta XYWH BBox coder. Following the practice in `R-CNN <https://arxiv.org/abs/1311.2524>`_, this coder encodes bbox (x1, y1, x2, y2) into de

Detection block in YOLO neck. Let out_channels = n, the DetectionBlock contains: Six ConvLayers, 1 Conv2D Layer and 1 YoloLayer. The firs

EfficientNet backbone. Args: arch (str): Architecture of efficientnet. Defaults to b0. out_indices (Sequence[int]): Output from w

EMAHook using exponential momentum strategy. Args: total_iter (int): The total number of iterations of EMA momentum. Defaults

Anchor-free head used in `FCOS <https://arxiv.org/abs/1904.01355>`_. The FCOS head does not use anchor boxes. Instead bounding boxes are pred