Returns the URL to the model trained using the given config Args: config_path (str): config file name relative to detectron2's "conf

Returns a config object for a model in model zoo. Args: config_path (str): config file name relative to detectron2's "configs/"

Args: image_size (tuple): height, width Returns: crop_size (tuple): height, width in absolute pixels

Load LVIS metadata. Args: dataset_name (str): LVIS dataset name without the split name (e.g., "lvis_v0.5"). Returns: di

Compute the output size given input size and target short edge length.

Get relative positional embeddings according to the relative positions of query and key sizes. Args: q_size (int): size of qu

Return the list of external input/output of sub-graph, each element is tuple of the name and corresponding version in predict_net. exter

Generalized Intersection over Union Loss (Hamid Rezatofighi et. al) https://arxiv.org/abs/1902.09630 Gradient-friendly IoU loss with an a

Args: predictions: return values of :meth:`forward()`. proposals (list[Instances]): proposals that match the features

Add object_descriptions and logit (if applicable) to detectron2's instances_to_coco_json

Dump an "Instances" object to a COCO-format json that's used for evaluation. Args: instances (Instances): img_id (int): the

The inverse is to rotate it back with expand, and crop to get the original shape.

Returns: dict[str -> (float, int)]: mapping from the name of each scalar to the most recent value and the iterati

Similar to :meth:`latest`, but the returned values are either the un-smoothed original latest value, or a median of the given

Args: image_dir (str): path to the raw dataset. e.g., "~/cityscapes/leftImg8bit/train". gt_dir (str): path to the raw annotations

Args: image_dir (str): path to the raw dataset. e.g., "~/cityscapes/leftImg8bit/train". gt_dir (str): path to the raw annotations

Args: image_dir (str): path to the raw dataset. e.g., "~/coco/train2017". gt_dir (str): path to the raw annotations. e.g., "~/coc

Load a json file in LVIS's annotation format. Args: json_file (str): full path to the LVIS json annotation file. image_root

Created list of ResNet stages from pre-defined depth (one of 18, 34, 50, 101, 152). If it doesn't create the ResNet variant you need,

Makes a donut shaped binary mask.

Convert pred_mask_logits to estimated foreground probability masks while also extracting only the masks for the predicted classes in pred_ins

If gradient clipping is enabled through config options, wraps the existing optimizer type to become a new dynamically created class Optimizer

Returns: (H, W) array, a mask for all pixels that have a prediction

Restrict angles to the range of [-180, 180) degrees

Given two lists of boxes of size N and M, compute the intersection area between __all__ N x M pairs of boxes. The box order must be (xmin

recursively (post-order) update all modules with the target type and its subclasses, make a initialization/composition/inheritance/... via th

support sigmoid

[{"from": "user", "value": f"<img>{file_abs_path}</img>" + prefix,}, {"from": "assistant", "value": label}] <|im_start|>system You are a

Normalize, pad and batch the input images.

Process the pair of inputs and outputs. If they contain batches, the pairs can be consumed one-by-one using `zip`: .. code-b

Save a jit module and load it back. Similar to the `getExportImportCopy` function in torch/testing/

This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch, num_key_value_heads, seqlen, he

Rotates half the hidden dims of the input.

Visualizes predictions on frames of the input video. Args: video (cv2.VideoCapture): a :class:`VideoCapture` object, who

Visualizes predictions on frames of the input video. Args: video (cv2.VideoCapture): a :class:`VideoCapture` object, who

Save the graph as SVG format. Args: output_file (str): a SVG file inputs: optional inputs given to the model

Save the model as caffe2's protobuf format. It saves the following files: * "model.pb": definition of the graph. Can be

Return semantic segmentation predictions in the original resolution. The input images are often resized when entering semantic segmentor. Mo

Turn a list of integer scalars or integer Tensor scalars into a vector, in a way that's both traceable and scriptable. In tracing, `x` s

Split a full s3 path into the bucket name and path.

Return `num_samples` (or fewer, if not enough found) random samples from `labels` which is a mixture of positives & negatives. It will tr

Convert InstancesList to List[Instances]. The input `instances_list` can also be a List[Instances], in this case this method is a non

Serializes this instance to a JSON string.

Converts a string in a sequence of tokens. Args: text (`str`): The sequence to be encoded. a

Upgrade a config from its current version to a newer version. Args: cfg (CfgNode): to_version (int): defaults to the latest

Args: results (OrderedDict[dict]): task_name -> {metric -> score} Returns: bool: whether the verification succeeds or not

Partition into non-overlapping windows with padding if needed. Args: x (tensor): input tokens with [B, H, W, C]. window_size

Args: loss_dict (dict): dict of scalar losses data_time (float): time taken by the dataloader iteration p

Args: min_sizes: list of short-edge size to resize the image to max_size: maximum height or width of resized images

Args: backbone: either a backbone module or a mmdet config dict that defines a backbone. The backbone takes a 4D

NOTE: this interface is experimental. Args: num_keypoints (int): number of keypoints to predict loss_weight

NOTE: this interface is experimental. Args: loss_weight (float): multiplier of the loss vis_period (int): vi

Args: backbone: a backbone module, must follow detectron2's backbone interface proposal_generator: a module that gene

Args: center_sampling_radius: radius of the "center" of a groundtruth box, within which all anchor points are lab

NOTE: this interface is experimental. Args: input_shape: shapes (channels and stride) of the input features