A simple structure that contains basic shape specification about a tensor. It is often used as the auxiliary inputs/outputs of models, to

A wrapper around :class:`torch.nn.Conv2d` to support empty inputs and more features.

This structure stores a list of boxes as a Nx4 torch.Tensor. It supports some common methods about boxes (`area`, `clip`, `nonempty`, etc

This class represents a list of instances in an image. It stores the attributes of instances (e.g., boxes, masks, labels, scores) as "fields"

This structure stores a list of rotated boxes as a Nx5 torch.Tensor. It supports some common methods about boxes (`area`, `clip`, `nonemp

Region Proposal Network, introduced by the Faster R-CNN paper.

Generalized R-CNN. Any models that contains the following three components: 1. Per-image feature extraction (aka backbone) 2. Region prop

A head with several 3x3 conv layers (each followed by norm & relu) and several fc layers (each followed by relu).

A mask head with several conv layers, plus an upsample layer (with `ConvTranspose2d`).

It's "standard" in a sense that there is no ROI transform sharing or feature sharing between tasks. The cropped rois go to separate branc

A timer which computes the time elapsed since the start/reset of the timer.

The box-to-box transform defined in R-CNN. The transformation is parameterized by 4 deltas: (dx, dy, dw, dh). The transformation scales the b

The user-facing class that provides metric storage functionalities. In the future we may add support for storing / logging other types of da

For a set of image sizes and feature maps, computes a set of anchors.

Region of interest feature map pooler that supports pooling from one or more feature maps.

This method returns a copy of this image, rotated the given number of degrees counter clockwise around its center.

This class stores the segmentation masks for all objects in one image, in the form of bitmaps. Attributes: tensor: bool Tensor o

Structure that holds a list of images (of possibly varying sizes) as a single tensor. This works by padding the images to the same size,

A transform that does nothing.

Same as :class:`Checkpointer`, but is able to handle models in detectron & cvpods model zoo, and apply conversions for legacy models.

This class assigns to each predicted "element" (e.g., a box) a ground-truth element. Each predicted element will have exactly zero or one mat

The registry that provides name -> object mapping, to support third-party users' custom modules. To create a registry (e.g. a backbone re

This class stores the segmentation masks for all objects in one image, in the form of polygons. Attributes: polygons: list[list[ndar

Creates a qualifier with the given name. test_func should be a callable object (e.g., lambda) that takes in as input an annotation objec

Two linear layers for predicting Fast R-CNN outputs: (1) proposal-to-detection box regression deltas (2) classification scores

Composes several transforms together.

Perform crop operations on images.

A class that stores information about outputs of a Fast R-CNN head.

This module implements simple decoder head for Semantic Segmentation. It creats decoder on top of the dynamic backbone.

Main class for semantic segmentation architectures.

Depthwise seperable convolution block.

For a set of image sizes and feature maps, computes a set of shifts.

Transforms pixel colors with PIL enhance functions.

The box-to-box transform defined in Rotated R-CNN. The transformation is parameterized by 5 deltas: (dx, dy, dw, dh, da). The transformation

A class to hold ground truth or predictions data in an easy to work with format. Note that any time they appear, bounding boxes are [x, y, wi

Implement Decoupled SOLO. See: https://arxiv.org/pdf/1512.02325.pdf for more details.

Compute default box coordinates (xmin, ymin, xmax, ymax) for each feature map. Returns: default_boxes (list['Boxes']): a list of 'Boxes' e

This module implements Dynamic Network for Semantic Segmentation.

Implement EfficientDet(https://arxiv.org/abs/1911.09070). See: https://arxiv.org/pdf/1911.09070.pdf for more details.

Write scalars to a json file. It saves scalars as one json per line (instead of a big json) for easy parsing. Examples parsing such a j

The ROIHeads in a typical "C4" R-CNN model, where the box and mask head share the cropping and the per-region feature computation by a Re

Implement SOLO: Segmenting Objects by Locations. See: https://arxiv.org/pdf/1512.02325.pdf for more details.

Implement SSD (https://arxiv.org/abs/1512.02325). See: https://arxiv.org/pdf/1512.02325.pdf for more details.

A semantic segmentation head described in :paper:`PanopticFPN`. It takes FPN features as input and merges information from all levels of

A point head multi-layer perceptron which we model with conv1d layers with kernel 1. The head takes both fine-grained and coarse prediction f

Implement the Swish activation function. See: https://arxiv.org/abs/1710.05941 for more details.

Stores all the information necessary to calculate the AP for one IoU and one class. Note: I type annotated this because why not.

Stores an APDataObject for each class in the dataset.

A mask head with fully connected layers. Given pooled features it first reduces channels and spatial dimensions with conv layers and then use

Print **common** metrics to the terminal, including iteration time, ETA, memory, all losses, and the learning rate. To print something d

A wrapper around :class:`torch.nn.ConvTranspose2d` to support zero-size tensor.

Implement test-time augmentation for detection data. It is a callable which takes a dataset dict from a detection dataset, and returns a

The head used in FCN for Semantic Segmentation. See: https://arxiv.org/abs/1605.06211 for more details.

Attribute: polygons (list[ndarray]): list[ndarray]: polygons for this mask. Each ndarray has format [x, y, x, y, ...]

Stores keypoint annotation data. GT Instances have a `gt_keypoints` property containing the x,y location and visibility flag of each keypoint

The RoI heads class for PointRend instance segmentation models. In this class we redefine the mask head of `StandardROIHeads` leaving all ot

Resize the image to a target size.

Implement RetinaNet (https://arxiv.org/abs/1708.02002).

r""" Single head used in CenterNet Head.

Maintain a list of transform operations which will be applied in sequence. Attributes: transforms (list[Transform])

Some errors are fixed by changing false positives to true positives. The issue with fixing these errors naively is that you might have mu

Enum of different ways to represent a box. Attributes: XYXY_ABS: (x0, y0, x1, y1) in absolute floating points coordinates.

Similar to nn.ParameterList, but for buffers

Evaluate object proposal, instance detection/segmentation, keypoint detection outputs using COCO's metrics and APIs.

r""" Implement CenterNet (https://arxiv.org/abs/1904.07850).

A wrapper around :class:`torch.nn.Conv2d` to support "SAME" padding mode and more features.

Implement FCOS (https://arxiv.org/abs/1708.02002).

The head used in FCOS for object classification and box regression. It has two subnets for the two tasks, with a common structure but separat

This module implements Feature Pyramid Network. It creates pyramid features built on top of some input feature maps.

A GeneralizedRCNN with test-time augmentation enabled. Its :meth:`__call__` method has the same interface as :meth:`GeneralizedRCNN.forward`.