recover the text from the tokens. this function is used for create caption of the visualzation.

sample two list with same length.

save the activation for one epoch.

save the activation for one epoch.

swap the subject in smpl and smal

swap the subject in smpl and smal

encode text with clip. inputs: ------- raw_text list (batch_size length) of strings with input text prom

encode text with clip. inputs: ------- raw_text list (batch_size length) of strings with input text prom

encode text with clip. inputs: ------- raw_text list (batch_size length) of strings with input text prom

visualize the reconstruction inputs: ------- rec_motions : [batch_size, num_frames, num_joints, ] the reconstructed motions

visualize the reconstruction inputs: ------- rec_motions : [batch_size, num_frames, num_joints, ] the reconstructed motions

the collate function for the smal motion dataset

the collate function for the smpl motion dataset

compute the offset from the joints inputs: ------- joints: np.ndarray the joints position, shape: (num_joints, 3)

evaluate the model for one full epoch.

find the template smal mesh.

forward from smal only dataset.

only forward for mm evaluation. no caption is needed.

gather the vertices from the obj file paths

load beta from json file with the image path keyword.

get the motion name from the animal name

get with_tail information from json file with img_keyword.

initialize betas_logscale, scale, theta, trans. only scale with one. others are with zero. inputs: ------- device: torch.device

Determine if a given vertex is in an n x 3 numpy array of vertices. Parameters: - query_vertex (numpy array): Array representing the ver

load scale from the template

log the losses to the SummaryWriter

log the losses to the SummaryWriter

main function for generate in distribution mean and cov

main function

main function for generate in distribution mean and cov

the main function for evaluate the mm for ood of model

the main function for load the motion, model, and generate metric results

main function

main function

the main function for evaluate the mm for ood of model

main function for training the mapping between SMPL and SMAL

main function for training the mapping between SMPL and SMAL

main function for procssing skeleton data

the main function to convert vertices to joints

take some time to figure out the index.

parse args

parser the args for training motion clip.

plot the obj file

process the joints to 6d representation

process the SMAL obj file

convert the internal data to json writable format

train one epoch inputs: ------- step the global current step

forward SMAL model and save the result to save_path

visualize the optimization results of the given animal

visualize the results

visualize the results

get item from the dataset

get item from the dataset

get item from the dataset

gmotion is short for generated motion

gmotion is short for generated motion

gmotion is short for generated motion

inputs: ------- list_steps the list of steps list_lambdas the list of lambdas