Prepare statistics for normalization

Prepare the statistics for normalization

Prepare statistics for normalization

Prepare the statistics of the dataset

Prepare the statistics for normalization

Prepare the statistics of the dataset

Save model and optimizer state.

Save model and optimizer state.

Convert original smplx parameters to transformed smplx parameters Args: sampled_trans: sampled valid position sample

Create a beta schedule that discretizes the given alpha_t_bar function, which defines the cumulative product of (1-beta) over time from t = [

Numpy implementation of the Frechet Distance. The Frechet distance between two multivariate Gaussians X_1 ~ N(mu_1, C_1) and X_2 ~ N(mu_2, C_2

Chamfer distance between two point clouds Reference: https://gist.github.com/sergeyprokudin/c4bf4059230da8db8256e36524993367 Parameters --

Compute the mean for the previous step, given a function cond_fn that computes the gradient of a conditional log probability with res

Compute what the p_mean_variance output would have been, should the model's score function be conditioned by cond_fn. See co

Convert raw smplx representation to pos representation (for HumanML3D dataset)

Create evaluator Args: cfg: configuration dict Returns: Evaluator

Create gaussian diffusion Args: cfg: configuration dict Return: Diffusion model

Create model according to the configuration Args: cfg: configuration dict Return: Model for prediction

Sample x_{t-1} from the model using DDIM. Same usage as p_sample().

Use DDIM to sample from the model and yield intermediate samples from each timestep of DDIM. Same usage as p_sample_loop_pro

Compute the log-likelihood of a Gaussian distribution discretizing to a given image. :param x: the target images. It is assumed that thi

Evaluate APD

Evaluate HumanML metrics

Evaluate physics, including non-collision and contact and dist

Convert image frames to gif, use PIL to implement the convertion. Args: frames: a image list or a image directory gif: save path

Convert joints to SMPL-X mesh verts Args: joints: joints position sequence (b, l, j * 3) joints_mask: sequence mask

Load humanise dataset Args: min_horizon: minimum horizon for motion sequence max_horizon: maximum horizon for motion sequence

Load scene model Args: model_name: Scene model name. model_dim: Scene model dimension. num_points: Number of points.

Main function Args: cfg: configuration dict

Main function Args: cfg: configuration dict

Main function Args: cfg: configuration dict

Compute the KL divergence between two gaussians. Shapes are automatically broadcasted, so batches can be compared to scalars, among othe

Normalize pose sequence Args: pose_seq: a numpy array of pose sequence Return: Normalized p

Normalize the contact map Args: contact: the contact map to be normalized Returns: normalized contact map

Normalize the motion Args: motion: motion data to be normalized Return: Normalized motion

Normalize the contact map Args: contact: the contact map to be normalized Returns: normalized contact map

Normalize the motion Args: motion: motion data to be normalized Return: Normalized motion

Optimize the body parameters with given joints Args: body_model: SMPLXLayer joints: <nframes, 22, 3> or <nframes, 66> in

Sample x_{t-1} from the model at the given timestep. :param model: the model to sample from. :param x: the current tensor at

Generate samples from the model and yield intermediate samples from each timestep of diffusion. Arguments are the same as p_

Process motion-condition pairs Args: motions: motion data scene_data: scene data num_points: number of points for ea

Diffuse the data for a given number of diffusion steps. In other words, sample from q(x_t | x_0). :param x_start: the initi

read XYZRGB point cloud from filename PLY file

Render meshes to videos Args: save_dir: directory to save videos meshes: meshes to render appendix_meshes: appendix mesh

Render meshes to videos Args: save_dir: directory to save videos meshes: meshes to render appendix_meshes: appendix mesh

Open-vocabulary?

Convert raw smplx representation to specific vector representation Args: smplx: A tuple containing motion sequence and betas dat

Create a list of timesteps to use from an original diffusion process, given the number of timesteps we want to take from equally-sized portio

Begin testing with this function Args: cfg: configuration dict

Generate a time string Args: Y, m, d, H, M, S: create year or not, default is True Return: time string

Begin training with this function Args: cfg: configuration dict

Update the reweighting using losses from a model. Sub-classes should override this method to update the reweighting using lo

Get a numpy array of weights, one per diffusion step. The weights needn't be normalized, but must be positive.