Args: preds_edges: with shape [b, c, h , w] edges: with shape [b, c, h, w] Returns: Edge losses

Moves the previous layer cache to the CPU

Extract features from images.

Extract features from images.

Write optical flow to file. If the flow is not quantized, it will be saved as a .flo file losslessly, otherwise a jpeg image which is lossy b

Placeholder of forward function.

imgs: [B, 3, H, W] of torch.float32. - mean: [0.48145466, 0.4578275, 0.40821073] - std: [0.26862954, 0.2613025

Placeholder of forward function.

Calls either forward_train or forward_test depending on whether return_loss is True. Note: - batch_size: N -

Forward function.

Forward function. Calls either forward_train or forward_test depending on whether return_loss=True. Note: - batc

Used for computing network FLOPs. See ``tools/get_flops.py``. Args: img (torch.Tensor): Input image. Returns:

x: [N, C] Float[Tensor, "Np Cp"], query_points: [N, 3] Float[Tensor, "Np 3"]

Forward pass of the latent points generation. Args: image (torch.Tensor): Input image tensor of shape [B, C_img, H_img, W

Forward pass of the latent points generation. Args: image (torch.Tensor): Input image tensor of shape [B, C_img, H_img, W

Args: imgs (List[Tensor]): the outer list indicates test-time augmentations and inner Tensor should have a shape

Args: imgs (List[Tensor]): the outer list indicates test-time augmentations and inner Tensor should have a shape

Defines the computation performed at every call when testing.

Defines the computation performed at training. Note: batch_size: N num_keypoints: K num_img_channel: C

Forward function for training. Args: img (Tensor): Input images. img_metas (list[dict]): List of image info dict wher

Forward function for training. Args: img (Tensor): Input images. img_metas (list[dict]): List of image info dict wher

Defines the computation performed at training. Note: batch_size: N num_keypoints: K num_img_channel: C

Positionally encode points that are not normalized to [0,1].

r"""Create an operator that blurs a tensor using a Gaussian filter. .. image:: _static/img/gaussian_blur2d.png The operator smooths the give

embed_dim: output dimension for each position pos: a list of positions to be encoded: size (M,) out: (M, D)

Calculate accuracy for keypoint loss. Note: - batch_size: N - num_keypoints: K Args: output (tor

Helper function to get activation function from string. Args: act_fn (str): Name of activation function. Returns: nn.Module:

this will perform the index select, cat the tensors, and provide the attn_bias from cache

this will perform the index select, cat the tensors, and provide the attn_bias from cache

using skinning to find lower body vertices.

Given image size, fov and principal point coordinates, return K the camera parameter matrix

Makes broadcastable attention and causal masks so that future and masked tokens are ignored. Arguments: attention_mask (

A helpfer function to get the keypoint indices of the face from the model config. Args: model_cfg (Config): pose model config. R

Return mesh faces. Note: F: number of faces Returns: faces: np.ndarray([F, 3]), mesh faces

Returns: The rank of the current process within the global process group.

Returns: The number of processes in the process group