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

Extract features from images.

flip tensor base on flip_direction. Args: src_tensor (Tensor): input feature map, shape (B, C, H, W). flip_direction (str): The f

Forward features from the upstream network. Args: feats (tuple[Tensor]): Features from the upstream network, each is

Forward function for `Transformer`. Args: x (Tensor): Input query with shape [bs, c, h, w] where c = embed_dims.

r"""Generate 2D gaussian radius. This function is modified from the `official github repo <https://github.com/princeton-vl/CornerNet-Lite/blo

Generate base anchors. Returns: list(torch.Tensor): Base anchors of a feature grid in multiple \ feature levels.

Base on the previous bboxes and regression values, we compute the regressed bboxes and generate the grids on the bboxes. :param reg:

Generate base anchors of a single level. Args: base_size (int | float): Basic size of an anchor. scales (torch.Tensor

Get custom accuracy w.r.t. cls_score and labels. Args: cls_score (torch.Tensor): The prediction with shape (N, C + 2).

Get squares according to feature map sizes and guided anchors. Args: featmap_sizes (list[tuple]): Multi-level feature map sizes.

Get squares according to feature map sizes and guided anchors. Args: featmap_sizes (list[tuple]): Multi-level feature map sizes.

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

preds: Bs*C*W*H

get_bboxes() is implemented in StageCascadeRPNHead.

Get class names of a dataset.

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

Get ``num_points`` most uncertain points during test. Args: mask_pred (Tensor): A tensor of shape (num_rois, num_classes,

Get sampled approxs and inside flags according to feature map sizes. Args: featmap_sizes (list[tuple]): Multi-level feature map s

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

Compute corresponding GT box and classification targets for proposals. Args: proposals_list (list[list]): Multi level poi

Compute semantic segmentation targets for each image. Args: segm_pred (Tensor): Predicted semantic segmentation map

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

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

Init module weights.

Are points located in bboxes. Args: points (Tensor): Points, shape: (m, 2). bboxes (Tensor): Bounding boxes, shape: (n, 4). Retu

Concat multi-level feature maps by image. [feature_level0, feature_level1...] -> [feature_image0, feature_image1...] Convert the shape of eac

A combination of the func:``AnchorHead.loss`` and func:``SSDHead.loss``. When ``self.use_ohem == True``, it functions like ``SSDHead.

Compute loss of the head. Args: segm_pred (list[Tensor]): Predicted semantic segmentation map with shape (N, num_

Compute loss of the head. Args: mask_pred (list[Tensor]): Predicted prototypes with shape (num_classes, H, W).

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

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

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

nn.Module: the normalization layer named "norm2"

nn.Module: normalization layer after the second convolution layer

Change from point offset to point coordinate.

Test without augmentation.

padding has no effect on polygons`

Prepare results dict for pipeline.

Print time spent by CPU and GPU. Useful as a temporary context manager to find sweet spots of code suitable for async implementation.

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

Randomly sample an img_scale when ``multiscale_mode=='range'``. Args: img_scales (list[tuple]): Images scale range for sampling.

Refine bboxes through stages.

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

see :func:`BaseInstanceMasks.rescale`

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

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

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

Set random seed. Args: seed (int): Seed to be used. deterministic (bool): Whether to set the deterministic option for

Simple test for mask head without augmentation.

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

SSD Head Config.

convert polygon masks to bitmap masks.

See :func:`BaseInstanceMasks.to_tensor`.

YoloV3 Head Config.

L2 normalization layer. Args: n_dims (int): Number of dimensions to be normalized scale (float, optional): Defaults t

Bottleneck block for ResNeXt. If style is "pytorch", the stride-two layer is the 3x3 conv layer, if it is "caffe", the stride-two lay

str: the string of the module

Set attribute, i.e. self.name = value This reloading prevent the teacher model from being registered as a nn.Module. The teacher modu

Add the fc branch which consists of a sequential of conv layers.