tensor shape: [b, c, h, w] paddings: [2 x 2] shape list, the first row indicates up and down paddings the second row indicates left and r

Flip the feature maps and swap the channels. Args: feature_maps (list[Tensor]): Feature maps. flip_index (list[int] | None): Chan

Flip human joints horizontally. Note: - batch_size: N - num_keypoint: K Args: regression (np.ndarray([..., K, C])):

Used for computing network FLOPs. See ``tools/get_flops.py``. Args: img (torch.Tensor[N,C,imgH,imgW], or list|tuple of t

Used for computing network FLOPs.

A helpfer function to get the keypoint indices of left and right eyes from the model config. Args: model_cfg (Config): pose model con

Calculate the transformation matrix under the constraint of unbiased. Paper ref: Huang et al. The Devil is in the Details: Delving into Unbiased

This function computes the inverse perspective projection of a set of points given an estimated distance. Input: points (bs, N, 2): 2

Calculate the Area Under the Curve (3DAUC) computed for a range of 3DPCK thresholds. Paper ref: `Monocular 3D Human Pose Estimation In The Wi

Calculate the mean per-joint position error (MPJPE) and the error after rigid alignment with the ground truth (P-MPJPE). Note: - batc

Load file form http url, will download models if necessary. Ref:https://github.com/1adrianb/face-alignment/blob/master/face_alignment/utils.py

In distributed setting, this function only download checkpoint at local rank 0.

Get multi-label classification accuracy. Note: - batch size: N - label number: L Args: pred (np.ndarray[N, L, 2]): m

Calculate oks ious. Args: g: Ground truth keypoints. d: Detected keypoints. a_g: Area of the ground truth object.

Group keypoints into poses given heatmap and tag. Note: batch size: N number of keypoints: K heatmap heig

Calculate the pose accuracy of PCK for each individual keypoint and the averaged accuracy across all keypoints from heatmaps. Note: P

model_name: query model_name

Setup logging. Args: output: A file name or a directory to save log files. If None, log files will not be saved. If outp

Args: x: shape is (B, T, H, W, C) window_size (tuple[int]): window size Returns: windows: (B, num_windows_h, num_wind

This encodes bbox(x,y,w,h) into (center, scale) Args: x, y, w, h Returns: tuple: A tuple containing center and scale.

calculate the distance of points p1 and p2.

Get configurations for deconv layers.

Calculate Gram matrix. Args: x (torch.Tensor): Tensor with shape of (n, c, h, w). Returns: torch.Tensor: Gra

Parse the raw outputs (losses) of the network. Args: losses (dict): Raw output of the network, which usually contain

sort kpts and remove the repeated ones.

Transform the bbox format from x1y1x2y2 to xywh. Args: bbox_xyxy (np.ndarray): Bounding boxes (with scores), shaped (n, 4) or

Aggregate multi-scale outputs. Note: batch size: N keypoints num : K heatmap width: W heatmap height: H Args

Check if the weight(and bias) is all zero.

Creates the optimizer

Compute the generalized charbonnier loss of the difference tensor x All positions where mask == 0 are not taken into account x: a tensor

Compose the standard form extrinsic matrix from RT. Batched I/O.

A simple wrapper for 3x3 convolution with padding. Args: inplanes (int): Channel number of inputs. outplanes (int): Channel numbe

return: (3, 2)

Compute the faces, barycentric coordinates for the dynamic landmarks To do so, we first compute the rotation of the neck around the y-axis a

Compute the faces, barycentric coordinates for the dynamic landmarks To do so, we first compute the rotation of the neck around the y-axis

Flip human joints horizontally. Note: - num_keypoints: K Args: joints_3d (np.ndarray([K, 3])): Coordinates of keypoints.

Read an optical flow map. Args: flow_path (ndarray or str): Flow path. quantize (bool): whether to read quantized pair, if set to

r"""Function that returns Gaussian filter coefficients. Args: kernel_size: filter size. It should be odd and positive. sigma: gau