Args: text (str): class label position (tuple): a tuple of the x and y coordinates to place text on image.

Filter out images with none annotations or only crowd annotations (i.e., images without non-crowd annotations). A common training-time pr

For each feature map, select the `pre_nms_topk` highest scoring proposals, apply NMS, clip proposals, and remove small boxes. Return the `pos

Freeze the first several stages of the ResNet. Commonly used in fine-tuning. Args: freeze_at (int): number of ste

Return dict from versioned blob to list of (i, j), where i is index of consumer op, j is the index of input of that op.

Returns: dict: a dict which maps names (str) to data of the fields Modifying the returned dict will modify this instance

Return dict from versioned blob to (i, j), where i is index of producer op, j is the index of output of that op.

Return the device type ("cpu" or "gpu"/"cuda") of each (versioned) blob

List all registered datasets. Returns: list[str]

Load semantic segmentation datasets. All files under "gt_root" with "gt_ext" extension are treated as ground truth annotations and all files

This method is almost identical to :meth:`RetinaNet.losses`, with an extra "loss_centerness" in the returned dict.

Yields: A context within which all the events added to this storage will be prefixed by the name scope.

Normalize the str to lower case and make them float numbers if possible.

Args: dataset_dicts (list[dict]): list of dataset dicts. class_names (list[str]): list of class names (zero-indexed).

Add an `img_tensor` associated with `img_name`, to be shown on tensorboard. Args: img_name (str): The name of th

Args: config_path: relative config filename

Args: box_scores (N, 5): boxes in corner-form and probabilities. (Note here 5 == 4 + 1, i.e., 4-dim horizontal bo

Rename the op_id-th operator in predict_net, change it's output_id-th input's name to the new_name. It also does automatic re-route and c

Compute (fractional) per-image repeat factors based on category frequency. The repeat factor for an image is a function of the freque

Rotates half the hidden dims of the input.

Scale the box with horizontal and vertical scaling factors

Given a list of N Instances (for N images), each containing a `gt_classes` field, return a list of Instances that contain only instances with

Args: see documentation of :func:`paste_masks_in_image`.

Serializes this instance to a Python dictionary.

Updates the underlying post processor with the current `bos_token` and `eos_token`.

Args: bottom_up (Backbone): module representing the bottom up subnetwork. Must be a subclass of :class:`Backbone`

NOTE: this interface is experimental. Args: num_classes (int): number of foreground classes (i.e. background is not incl

Args: cfg (CfgNode): torch_model (nn.Module): the detectron2 model (meta_arch) to be converted.

Naive tracking heuristics to assign same color to the same instance, will update the internal state of tracked instances. Re

Execute inference on a list of inputs, using batch size = self.batch_size, instead of the length of the list. Inputs & outpu

Caffe2 implementation of preprocess_image, which is called inside each MetaArch's forward. It normalizes the input images, and the fi

Calculate proper im2col step size, which should be divisible by input_size and not larger than prefer_size. Meanwhile the step size s

Returns: bool: whether the arguments contain CfgNode and should be considered forwarded to from_config.

Args: boxes (list[Tensor]): per-image predicted boxes, each of shape Ri x 4 image_sizes (list[tuple]): list of image

Run `_decode_per_level_predictions` for all feature levels and concat the results.

Transform anchors into proposals by applying the predicted anchor deltas. Returns: proposals (list[Tensor]): A list of L

x: LNHiWi x C

Find proper places to draw text given a binary mask.

Apply EMA update to `self.name` using `value`. This is mainly used for loss normalizer. In Detectron1, loss is normalized by number

Evaluate the coco results using COCOEval API.

Forward logic of the box prediction branch. If `self.train_on_pred_boxes is True`, the function puts predicted boxes in the `prop

Args: features, targets: the same as in Same as in :meth:`ROIHeads.forward`. proposals (list[Instance

Use `from_config` to obtain explicit arguments. Returns: dict: arguments to be used for cls.__init__

Return a process group based on gloo backend, containing all the ranks The result is cached.

Return the learning rate warmup factor at a specific iteration. See :paper:`ImageNet in 1h` for more details. Args: method (str)

A context which ignores CUDA OOM exception from pytorch.