Download checkpoint if not cached, return local path Args: url: URL to checkpoint file cache_dir: Optional cache directo

Load pretrained model (Transformers-style). Args: pretrained_model_name_or_path: Model name or path to saved mod

Get maximum response location and value from simcc representations. Note: instance number: N num_keypoints: K heatmap hei

Encoding keypoints into SimCC labels with Gaussian Label Smoothing strategy.

Forward the network. The input is the feature map extracted by backbone and the output is the simcc representation. Args:

Mapping keypoint coordinates into SimCC space.

Automatically patch MMDetection to accept mmcv 2.2.0

Decode keypoints from outputs. Args: batch_outputs (Tensor | Tuple[Tensor]): The network outputs of a data batch

Draw keypoints on image. Args: image: Input image (BGR) keypoints_2d: 2D keypoints array [N, 2] scores: Confiden

Calculate the pose accuracy of PCK for each individual keypoint and the averaged accuracy across all keypoints from 3D SimCC. Note: -

Test RTMPose3D inference on a single image. Args: image_path: Path to input image device: Device to run inference on ('c

Compare detection with and without single_person mode. Args: image_path: Path to input image device: Device to run infer

Move model to device (PyTorch-style). Args: device: Device name (e.g., 'cuda:0', 'cpu') Ret

Convert config to dictionary.

Visualize keypoints on image. Args: image_path: Path to input image output_path: Path to save annotated image (optional)

Run inference on image (Transformers-style). Args: image: Input image as numpy array (H, W, 3) in BGR format

Run 3D pose estimation on a numpy array image Args: image: Numpy array (HWC, BGR format) bbox_thr: B

Initialize RTMPose3D model with configuration. Args: config: RTMPose3DConfig instance

Initialize RTMPose3D inference pipeline Args: model_size: Model size ('l' for large, 'x' for extra large)

Add predictions into data samples. Args: batch_pred_instances (List[InstanceData]): The predicted instances of th

Remove all cached checkpoints

Decode SimCC labels into 3D keypoints. Args: encoded (Tuple[np.ndarray, np.ndarray]): SimCC labels for x-axis, y-axis

Get the device model is running on.

Alias for __call__ (Transformers-style). Args: image: Input image **kwargs: Additional arguments pas

Load configuration from directory or model name. Args: pretrained_model_name_or_path: Path to directory or model

Calculate losses from a batch of inputs and data samples.

Loss Name. Returns: str: The name of this loss item.

Predict results from features. Args: feats (Tuple[Tensor] | List[Tuple[Tensor]]): The multi-stage features (or mu

Save configuration to directory. Args: save_directory: Directory to save config.json

Save model configuration to directory (Transformers-style). Note: Checkpoints are not saved, only configuration. Use