Transform inputs for decoder. Args: inputs (list[Tensor]): List of multi-level img features. Returns: Tensor

Classify each pixel with fc.

Convert a version string into a tuple of integers. This method is usually used for comparing two versions. For pre-release versions: alpha <

Get the root logger. The logger will be initialized if it has not been initialized. By default a StreamHandler will be added. If `log_file` i

Add prefix for dict. Args: inputs (dict): The input dict with str keys. prefix (str): The prefix to add. Returns: d

Apply element-wise weight and reduce loss. Args: loss (Tensor): Element-wise loss. weight (Tensor): Element-wise weights.

Apply element-wise weight and reduce loss. Args: loss (Tensor): Element-wise loss. weight (Tensor): Element-wise weights.

Forward function for plugins.

Args: imgs (List[Tensor]): the outer list indicates test-time augmentations and inner Tensor should have a shape

Inference with slide/whole style. Args: img (Tensor): The input image of shape (N, 3, H, W). img_meta (dict): Image i

nn.Module: the normalization layer named "norm2"

Build projection layer for key/query/value/out.

Get class weight for loss function. Args: class_weight (list[float] | str | None): If class_weight is a str, take it as a fil

Get class weight for loss function. Args: class_weight (list[float] | str | None): If class_weight is a str, take it as a fil

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

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.

Make each stage.

Make transition layer.

Calculate accuracy according to the prediction and target. Args: pred (torch.Tensor): The model prediction, shape (N, num_class, ...)

Build backbone.

Build loss.

Encode images with backbone and decode into a semantic segmentation map of the same size as input.

Calculate evaluation metrics Args: results (list[ndarray] | list[str]): List of prediction segmentation maps or list of predic

Placeholder for Forward function for training.

Forward function for training. Args: img (Tensor): Input images. img_metas (list[dict]): List of image info dict wher

make plugins for block. Args: in_channels (int): Input channels of plugin. plugins (list[dict]): List of plugins cfg

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.

Inference by sliding-window with overlap. If h_crop > h_img or w_crop > w_img, the small patch will be used to decode without padding. A

Forward function for feature maps before classifying each pixel with ``self.cls_seg`` fc. Args: inputs (list[Tensor]): Li

Freeze stages param and norm stats.

Freeze stages param and norm stats.

Sample from fine grained features. Args: prev_output (list[Tensor]): Prediction of previous decode head. points (Tens

Sample from fine grained features. Args: x (list[Tensor]): Feature pyramid from by neck or backbone. points (Tensor):