Assign boxes to either a ground truth boxes or a negative boxes.

Builder of box assigner.

Builder of box sampler.

Build shader.

Convert intrinsic matrix from ndc to screen. Args: K (Union[torch.Tensor, np.ndarray]): Input 4x4 intrinsic matrix, left mm d

Convert intrinsic matrix from screen to ndc. Args: K (Union[torch.Tensor, np.ndarray]): input intrinsic matrix. resolution (Union

If a key contains 'keypoints', and f'{key}_mask' is in self.keys(), invalid zeros will be inserted to the right places and f'{key}_mask'

Modified from pytorch3d and add `model_type` in __class__.__init__.

Convert euler angle to rotation matrixs. Args: euler_angle (Union[torch.Tensor, numpy.ndarray]): input shape should be (.

convert camera convention to axis direction and order.

Find camera translation that brings 3D joints S closest to 2D the corresponding joints_2d. Input: S: (B, 49, 3) 3D joint locations

Evaluate fitted parameters through loss computation. This function serves two purposes: 1) internally, for loss backpropagation 2) ext

Turn param tensors to dict.

Build camera param.

Get the transformation of points on the cropped image to the points on the original image.

Class method. Calculate the projective transformation matrix by default parameters. Args: **kwargs: parameters for the pr

Get color image resolution of an iPhone RGB camera by iPhone id. Args: iphone_id (int, optional): ID of an iPhone

Get extrinsics(cam2world) of an iPhone RGB camera by iPhone id. Args: iphone_id (int, optional): ID of an iPhone,

Read a folder of images as an array of (f * h * w * 3). Args: input_folder (str): folder of input images. resolution (Option

Initialize model weights.

Args: feat (torch.Tensor): [B, C, H, W] image features uv (torch.Tensor): [B, 2, N] uv coordinates in the image plane

Inverse function of sigmoid. Args: x (Tensor): The tensor to do the inverse. eps (float): EPS avoid numerical

Boolean of whether is perspective.

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

Calculate the Percentage of Correct Keypoints (3DPCK) w. or w/o rigid alignment. Paper ref: `Monocular 3D Human Pose Estimation In The Wild Us

Load camera name and parameters from a file.

Build decoders for each pose.

Build regressor for ExPose Head.

Loss function. Only outputs from the last feature level are used for computing losses by default. Args: all_cls_

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

Merge list into cfg_dict. Merge the dict parsed by MultipleKVAction into this cfg. Examples: >>> options = {'model.backb

Perform orthographic projection of 3D points using the camera parameters, return projected 2D points in image plane. Notes: batch siz

Convert image tensor to array.

Convert full pose tensor to pose dict. Args: full_pose (torch.Tensor): shape should be (..., 165) or (..., 55, 3)

Move the pixels lie on the edges inside the mask, then refine the rest points by searching the nearest pixel in the faces it should be

Update pix_to_face, bary_coords.

Pad zeros to a compressed keypoints array. Args: compressed_array (np.ndarray): A compressed keypoints array.

Check whether the key matches definition in HumanData.SUPPORTED_KEYS. Args: key (Any): Key in HumanData.

Check whether the value matches definition in HumanData.SUPPORTED_KEYS. Args: key (Any): Key in HumanData

Concat two values from two different HumanData. Args: key (Any): The common key of the two values. va

Get value defined by HumanData. This function will be called by self[key]. In keypoints_compressed mode, if the key contains 'keypoint

Args: *args: extra arguments for SMPL initialization. keypoint_src: source convention of keypoints. This convention

PARE parameters regressor head. This class is modified from. [PARE](hhttps://github.com/ mkocabas/PARE/blob/master/pare/models/head/p

str: a "nice" summary string describing this module

str: a "nice" summary string describing this module

Decode an RGB image from an encoded byte array.

Remove zero-padding from a full-size keypoints array. Args: zero_pad_array (np.ndarray): A keypoints array in ful

For older versions of HumanData, call this method to apply missing values (also attributes). Note: 1. Older HumanData doesn't

str: the string of the module

Return a zero mat in list format. Args: n (int, optional): Length of the edge. Defaults to 3. Returns:

Loss computation. Notes: B: batch size K: number of keypoints D: shape dimension Args:

Compute relative loss change. If relative change is small enough, we can apply early stop to accelerate the optimization. (1) When one of

Construct the in-plane rotation matrix. Args: rot (float): Rotation angle (degree). size (int): The size of the rotation matrix.

A helper function to expand the betas's first dim to match batch size such that the same beta parameters can be used for all frames in a

Flip axis_angle horizontally. Args: r (np.ndarray([3])) Returns: f_flipped

To calculate the affine matrix, three pairs of points are required. This function is used to get the 3rd point, given 2D points a & b. The 3r

Return an activation function given a string.

Return an activation function given a string.

Transform outputs from the last decoder layer into bbox predictions for each image. Args: cls_score (Tensor): Box score l

get deconv padding, output padding according to kernel size.

mlp layers for shape regression.

make convolution layers.

merge dict `a` into dict `b` (non-inplace). values in `a` will overwrite `b`. copy first to avoid inplace modification. Args:

Cauculate the Area Under the Curve (AUC) computed for a range of 3DPCK thresholds. Args: keypoint_results (list): Keypoint

Cauculate the Area Under the Curve (AUC) computed for a range of 3DPCK thresholds. Args: keypoint_results (list): Keypoint

Cauculate Percentage of Correct Keypoints (3DPCK) w. or w/o Procrustes alignment. Args: keypoint_results (list): Keypoint

Cauculate Percentage of Correct Keypoints (3DPCK) w. or w/o Procrustes alignment. Args: keypoint_results (list): Keypoint

Cauculate mean per joint position error (MPJPE) or its variants PA- MPJPE. Report mean per joint position error (MPJPE) and mean per

Cauculate mean per joint position error (MPJPE) or its variants PA- MPJPE. Report mean per joint position error (MPJPE) and mean per