Functions — github.com/Walter0807/MotionBERT

CodeHub MCPcopy Create free account

hub / github.com/Walter0807/MotionBERT / functions

Functions186 in github.com/Walter0807/MotionBERT

⨍ Functions 186◇ Types & classes 30

↓ 34 callersMethodupdate

(self, val, n=1)

lib/utils/learning.py:19

↓ 12 callersFunctionflip_data

horizontal flip data: [N, F, 17, D] or [F, 17, D]. X (horizontal coordinate) is the first channel in D. Return result: same

lib/utils/utils_data.py:54

↓ 9 callersFunctionrgb2rgba

lib/utils/vismo.py:75

↓ 6 callersFunctioncrop_scale

Motion: [(M), T, 17, 3]. Normalize to [-1, 1]

lib/utils/utils_data.py:7

↓ 6 callersFunctionget_config

lib/utils/tools.py:46

↓ 6 callersFunctionload_backbone

lib/utils/learning.py:79

↓ 6 callersFunctionsplit_clips

(vid_list, n_frames, data_stride)

lib/utils/utils_data.py:91

↓ 5 callersFunctionread_pkl

lib/utils/tools.py:65

↓ 4 callersMethodforward_spatial

(self, q, k, v)

lib/model/DSTformer.py:178

↓ 4 callersFunctionget_angles

Input: (N, T, 17, 3) Output: (N, T, 16)

lib/model/loss.py:148

↓ 4 callersFunctionpartial_train_layers

Train partial layers of a given model.

lib/utils/learning.py:69

↓ 4 callersFunctionrender_and_save

(motion_input, save_path, keep_imgs=False, fps=25, color="#F96706#FB8D43#FDB381", with_conf=False, draw_face=F

lib/utils/vismo.py:17

↓ 4 callersMethodreshape_T

(self, x, seqlen=1, inverse=False)

lib/model/DSTformer.py:152

↓ 4 callersFunctionvalidate

(test_loader, model, criterion, dataset_name='h36m')

train_mesh.py:51

↓ 3 callersMethod__init__

(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.)

lib/model/DSTformer.py:70

↓ 3 callersFunctionbounding_box

lib/utils/vismo.py:344

↓ 3 callersMethodforward_temporal

(self, q, k, v, seqlen=8)

lib/model/DSTformer.py:188

↓ 3 callersFunctionget_limb_lens

Input: (N, T, 17, 3) Output: (N, T, 16)

lib/model/loss.py:98

↓ 3 callersMethodget_split_id

lib/data/datareader_h36m.py:100

↓ 3 callersFunctionloss_mpjpe

Mean per-joint position error (i.e. mean Euclidean distance), often referred to as "Protocol #1" in many papers.

lib/model/loss.py:56

↓ 3 callersFunctionsave_checkpoint

(chk_path, epoch, lr, optimizer, model_pos, min_loss)

↓ 3 callersFunctiontrain_epoch

(args, opts, model, train_loader, losses_train, losses_dict, mpjpes, mpves, criterion, optimizer, batch_time,

train_mesh.py:165

↓ 3 callersFunctiontrain_epoch

(args, model_pos, train_loader, losses, optimizer, has_3d, has_gt)

↓ 3 callersFunctiontrunc_normal_

r"""Fills the input Tensor with values drawn from a truncated normal distribution. The values are effectively drawn from the normal distributi

lib/model/DSTformer.py:48

↓ 2 callersMethod__init__

(self, backbone, dim_rep=512, num_classes=60, dropout_ratio=0., version='class', hidden_dim=2048, num_joints=1

lib/model/model_action.py:51

↓ 2 callersFunctionaccuracy

Computes the accuracy over the k top predictions for the specified values of k

lib/utils/learning.py:25

↓ 2 callersFunctionbatch_rodrigues

lib/utils/utils_mesh.py:8

↓ 2 callersFunctioncompute_error

(output, target)

lib/utils/utils_mesh.py:357

↓ 2 callersFunctionensure_dir

create path by first checking its existence, :param paths: path :return:

lib/utils/tools.py:56

↓ 2 callersFunctionevaluate

(args, model_pos, test_loader, datareader)

↓ 2 callersFunctionextract_feats

(dataloader_x, model)

train_action_1shot.py:44

↓ 2 callersFunctionflip_thetas_batch

Flip thetas in batch. Parameters ---------- thetas : numpy.array Joints in shape (N, F, num_thetas*3) theta_pairs : list

lib/utils/utils_mesh.py:486

↓ 2 callersFunctionget_img_from_fig

lib/utils/vismo.py:65

↓ 2 callersMethodget_representation

lib/model/DSTformer.py:360

↓ 2 callersMethodget_split_id

lib/data/datareader_mesh.py:42

↓ 2 callersFunctionjoints2image

(joints_position, colors, transparency=False, H=1000, W=1000, nr_joints=49, imtype=np.uint8, grayscale=False,

lib/utils/vismo.py:89

↓ 2 callersFunctionload_pretrained_weights

Load pretrianed weights to model Incompatible layers (unmatched in name or size) will be ignored Args: - model (nn.Module): network model,

lib/utils/learning.py:39

↓ 2 callersFunctionloss_angle

Input: (N, T, 17, 3), (N, T, 17, 3)

lib/model/loss.py:184

↓ 2 callersFunctionloss_angle_velocity

Mean per-angle velocity error (i.e. mean Euclidean distance of the 1st derivative)

lib/model/loss.py:192

↓ 2 callersFunctionloss_limb_gt

Input: (N, T, 17, 3), (N, T, 17, 3)

lib/model/loss.py:125

↓ 2 callersFunctionloss_limb_var

Input: (N, T, 17, 3)

lib/model/loss.py:114

↓ 2 callersFunctionloss_velocity

Mean per-joint velocity error (i.e. mean Euclidean distance of the 1st derivative)

lib/model/loss.py:133

↓ 2 callersFunctionn_mpjpe

Normalized MPJPE (scale only), adapted from: https://github.com/hrhodin/UnsupervisedGeometryAwareRepresentationLearning/blob/master/losses/po

lib/model/loss.py:79

↓ 2 callersFunctionnorm_cdf

lib/model/DSTformer.py:15

↓ 2 callersFunctionpixel2world_vis_motion

(motion, dim=2, is_tensor=False)

lib/utils/vismo.py:43

↓ 2 callersFunctionrandom_move

(data_numpy, angle_range=[-10., 10.], scale_range=[0.9, 1.1],

lib/data/dataset_action.py:76

↓ 2 callersMethodread_2d

lib/data/datareader_mesh.py:19

↓ 2 callersMethodread_3d

lib/data/datareader_h36m.py:60

↓ 2 callersFunctionresample

(ori_len, target_len, replay=False, randomness=True)

lib/utils/utils_data.py:68

↓ 2 callersFunctionrigid_align

lib/utils/utils_mesh.py:352

↓ 2 callersFunctionsave_clips

(subset_name, root_path, train_data, train_labels)

tools/convert_h36m.py:12

↓ 2 callersFunctionvalidate

(anchor_loader, test_loader, model)

train_action_1shot.py:58

↓ 2 callersFunctionvalidate

(test_loader, model, criterion)

train_action.py:41

↓ 1 callersMethod__init__

(self, args, dim_rep=512, num_joints=17, hidden_dim=2048, dropout_ratio=0.)

lib/model/model_mesh.py:10

↓ 1 callersMethod__init__

(self, args, subset_list, data_split)

lib/data/dataset_motion_3d.py:14

↓ 1 callersMethod__init__

(self, data_path, data_split, n_frames=243, random_move=True, scale_range=[1,1])

lib/data/dataset_action.py:170

↓ 1 callersMethod__init__

(self, n_frames=81, data_stride=27, flip=True, valid_threshold=0.0, scale_range=[0.25, 1])

lib/data/dataset_motion_2d.py:124

↓ 1 callersFunction_no_grad_trunc_normal_

(tensor, mean, std, a, b)

lib/model/DSTformer.py:12

↓ 1 callersMethodadd_mask

motion_2d: (N,T,17,3)

lib/data/augmentation.py:67

↓ 1 callersMethodadd_noise

(self, motion_2d)

lib/data/augmentation.py:29

↓ 1 callersMethodaugment2D

(self, motion_2d, mask=False, noise=False)

lib/data/augmentation.py:76

↓ 1 callersMethodaugment3D

(self, motion_3d)

lib/data/augmentation.py:94

↓ 1 callersFunctioncoco2h36m

Input: x (M x T x V x C) COCO: {0-nose 1-Leye 2-Reye 3-Lear 4Rear 5-Lsho 6-Rsho 7-Lelb 8-Relb 9-Lwri 10-Rwri 11-Lhip 12-Rhip

lib/data/dataset_action.py:31

↓ 1 callersFunctioncrop_scale_3d

Motion: [T, 17, 3]. (x, y, z) Normalize to [-1, 1] Z is relative to the first frame's root.

lib/utils/utils_data.py:31

↓ 1 callersMethoddenormalize

(self, test_data)

lib/data/datareader_h36m.py:125

↓ 1 callersMethoddis2conf

(self, dis, a, b, m, s)

lib/data/augmentation.py:22

↓ 1 callersFunctiondrop_path

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks). This is the same as the DropConnect impl I created for Ef

lib/model/drop.py:17

↓ 1 callersFunctionestimate_translation_np

This function is borrowed from https://github.com/nkolot/SPIN/utils/geometry.py Find camera translation that brings 3D joints S closest to 2

lib/utils/utils_mesh.py:222

↓ 1 callersFunctionevaluate_mesh

lib/utils/utils_mesh.py:395

↓ 1 callersFunctionflip_thetas

Flip thetas. Parameters ---------- thetas : numpy.ndarray Joints in shape (F, num_thetas, 3) theta_pairs

lib/utils/utils_mesh.py:458

↓ 1 callersMethodforward

(self, x, return_rep=False)

lib/model/DSTformer.py:329

↓ 1 callersMethodforward_coupling

(self, q, k, v, seqlen=8)

lib/model/DSTformer.py:163

↓ 1 callersMethodget_hw

lib/data/datareader_h36m.py:109

↓ 1 callersMethodget_sliced_data

lib/data/datareader_h36m.py:116

↓ 1 callersFunctionget_smpl_faces

lib/utils/utils_smpl.py:86

↓ 1 callersFunctionhalpe2h36m

Input: x (T x V x C) //Halpe 26 body keypoints {0, "Nose"}, {1, "LEye"}, {2, "REye"}, {3, "LEar"}, {4, "REa

lib/data/dataset_wild.py:15

↓ 1 callersFunctionhex2rgb

(hex, number_of_colors=3)

lib/utils/vismo.py:78

↓ 1 callersFunctionhuman_tracking

lib/data/dataset_action.py:114

↓ 1 callersFunctionjson2pose

tools/convert_insta.py:25

↓ 1 callersFunctionload_motion

tools/convert_insta.py:53

↓ 1 callersMethodlog

lib/utils/tools.py:15

↓ 1 callersFunctionloss_2d_weighted

(predicted, target, conf)

lib/model/loss.py:72

↓ 1 callersFunctionmake_cam

Input: x (M x T x V x C) img_shape (height, width)

lib/data/dataset_action.py:19

↓ 1 callersFunctionmotion2video

(motion, save_path, colors, h=512, w=512, bg_color=(255, 255, 255), transparency=False, motion_tgt=None, fps=2

lib/utils/vismo.py:213

↓ 1 callersFunctionmotion2video_3d

(motion, save_path, fps=25, keep_imgs = False)

lib/utils/vismo.py:246

↓ 1 callersFunctionmotion2video_mesh

(motion, save_path, fps=25, keep_imgs = False, draw_face=True)

lib/utils/vismo.py:287

↓ 1 callersFunctionmpjpe

Mean per-joint position error (i.e. mean Euclidean distance), often referred to as "Protocol #1" in many papers.

lib/model/loss.py:8

↓ 1 callersFunctionp_mpjpe

Pose error: MPJPE after rigid alignment (scale, rotation, and translation), often referred to as "Protocol #2" in many papers.

lib/model/loss.py:16

↓ 1 callersFunctionparse_args

train_mesh.py:33

↓ 1 callersFunctionparse_args

infer_wild_mesh.py:28

↓ 1 callersFunctionparse_args

↓ 1 callersFunctionparse_args

train_action_1shot.py:32

↓ 1 callersFunctionparse_args

infer_wild.py:15

↓ 1 callersFunctionparse_args

train_action.py:29

↓ 1 callersFunctionparse_args

tools/convert_insta.py:16

↓ 1 callersFunctionposetrack2h36m

Input: x (T x V x C) PoseTrack keypoints = [ 'nose', 'head_bottom',

lib/data/dataset_motion_2d.py:14

↓ 1 callersFunctionquat2mat

This function is borrowed from https://github.com/MandyMo/pytorch_HMR/blob/master/src/util.py#L50 Convert quaternion coefficients to rotatio

lib/utils/utils_mesh.py:23

↓ 1 callersFunctionquaternion_to_angle_axis

This function is borrowed from https://github.com/kornia/kornia Convert quaternion vector to angle axis of rotation. Adapted from ceres

lib/utils/utils_mesh.py:86

↓ 1 callersMethodread_2d

lib/data/datareader_h36m.py:25

↓ 1 callersMethodread_hw

lib/data/datareader_h36m.py:85

next →1–100 of 186, ranked by callers