Functions — github.com/anliyuan/Ultralight-Digital-Human

hub / github.com/anliyuan/Ultralight-Digital-Human / functions

Functions297 in github.com/anliyuan/Ultralight-Digital-Human

⨍ Functions 297◇ Types & classes 61

↓ 1 callersMethoddistance2bbox

(self, points, distance, max_shape=None)

data_utils/detect_face.py:35

↓ 1 callersMethoddistance2kps

(self, points, distance, max_shape=None)

data_utils/detect_face.py:46

↓ 1 callersFunctiondump_sample

(net: nn.Module, dataset, save_dir: str, epoch: int, device: torch.device)

train.py:150

↓ 1 callersFunctionencode_chunk

(model, ck_feat, offset, required_cache_size, subsampling_cache, elayers_output_cache, conformer_cnn_cache, de

data_utils/wenet/bin/recognize_wav_streaming.py:130

↓ 1 callersMethodendpoint_detected

(self)

data_utils/wenet_infer.py:447

↓ 1 callersFunctionextract_audio

(video_path: str, wav_path: str, sample_rate: int = 16000)

data_utils/process.py:20

↓ 1 callersFunctionextract_images

(video_path: str, mode: str)

data_utils/process.py:45

↓ 1 callersFunctionface_det

(img, model)

data_utils/get_landmark.py:14

↓ 1 callersFunctionfeature_extraction

(waveform, num_mel_bins=80, frame_length=25, frame_shift=10, dither=0.0, sample_rate=16000)

data_utils/wenet/bin/recognize_wav.py:107

↓ 1 callersFunctionfeature_extraction

(waveform, num_mel_bins=80, frame_length=25, frame_shift=10, dither=0.0, sample_rate=16000)

data_utils/wenet/bin/recognize_wav_streaming.py:112

↓ 1 callersFunctionforced_align

ctc forced alignment. Args: torch.Tensor ctc_probs: hidden state sequence, 2d tensor (T, D) torch.Tensor y: id sequence tensor 1d

data_utils/wenet/utils/ctc_util.py:16

↓ 1 callersMethodforward_chunk

Forward just one chunk Args: xs (torch.Tensor): chunk input offset (int): current offset in encoder output time stam

data_utils/wenet/transformer/encoder_streaming.py:169

↓ 1 callersMethodforward_one_step

Forward one step. This is only used for decoding. Args: memory: encoded memory, float32 (batch, maxlen_in, feat)

data_utils/wenet/transformer/decoder_streaming.py:131

↓ 1 callersFunctiongenerator_textgrid

(maxtime, lines, output)

data_utils/wenet/bin/alignment.py:35

↓ 1 callersFunctionget_args

()

data_utils/wenet/bin/train.py:37

↓ 1 callersFunctionget_args

()

data_utils/wenet/bin/recognize_wav.py:37

↓ 1 callersFunctionget_args

()

data_utils/wenet/bin/recognize_wav_streaming.py:42

↓ 1 callersFunctionget_args

()

data_utils/wenet/bin/recognize.py:33

↓ 1 callersFunctionget_args

()

data_utils/wenet/bin/export_jit.py:27

↓ 1 callersFunctionget_args

()

data_utils/wenet/bin/average_model.py:12

↓ 1 callersFunctionget_audio_feature

(wav_path: str, mode: str)

data_utils/process.py:65

↓ 1 callersFunctionget_frames_timestamp

(alignment)

data_utils/wenet/bin/alignment.py:53

↓ 1 callersFunctionget_labformat

(timestamp, subsample)

data_utils/wenet/bin/alignment.py:74

↓ 1 callersFunctionget_landmark

(video_path: str, landmarks_dir: str)

data_utils/process.py:89

↓ 1 callersMethodget_waveform_len

(self)

data_utils/FeaturePipeline.py:65

↓ 1 callersFunctioninsert_blank

Insert blank token between every two label token.

data_utils/wenet/utils/ctc_util.py:7

↓ 1 callersFunctionload_model

(checkpoint_path: str, mode: str, device: torch.device)

inference.py:58

↓ 1 callersFunctionmain

(arg)

dihuman_run.py:276

↓ 1 callersFunctionmain

()

train.py:164

↓ 1 callersFunctionmain

()

pth2onnx.py:38

↓ 1 callersFunctionmain

()

data_utils/process.py:112

↓ 1 callersFunctionmain

()

data_utils/hubert.py:77

↓ 1 callersFunctionmain

()

data_utils/wenet/bin/train.py:108

↓ 1 callersFunctionmain

()

data_utils/wenet/bin/recognize_wav.py:125

↓ 1 callersFunctionmain

()

data_utils/wenet/bin/recognize_wav_streaming.py:213

↓ 1 callersFunctionmain

()

data_utils/wenet/bin/recognize.py:104

↓ 1 callersFunctionmain

()

data_utils/wenet/bin/export_jit.py:39

↓ 1 callersFunctionmain

()

data_utils/wenet/bin/average_model.py:39

↓ 1 callersFunctionmake_even_first_dim

让 T 维变成偶数。原实现是直接丢一帧，这里改成 pad 一帧 0，避免丢失最后一帧信息。

data_utils/hubert.py:68

↓ 1 callersFunctionmerge_audio

(video_path: str, audio_path: str, output_path: str)

inference.py:67

↓ 1 callersMethodnext

(self)

inference.py:94

↓ 1 callersFunctionparse_args

()

dihuman_run.py:35

↓ 1 callersFunctionparse_args

()

train.py:25

↓ 1 callersFunctionparse_args

()

inference.py:34

↓ 1 callersMethodprocess

(self, audio_frame: np.ndarray)

dihuman_run.py:228

↓ 1 callersMethodreset

收到首段有效音频时清空缓存。

dihuman_run.py:147

↓ 1 callersMethodresize_image

(self, srcimg)

data_utils/detect_face.py:17

↓ 1 callersFunctionresume_if_any

返回起始 epoch。空路径表示从头训练。

train.py:87

↓ 1 callersFunctionrun

(args, device: Optional[torch.device] = None)

inference.py:138

↓ 1 callersFunctionselect_fourcc

根据输出文件扩展名选择合适的视频编码器。

inference.py:50

↓ 1 callersFunctiontrain_one_epoch

( net: nn.Module, loader: DataLoader, optimizer: optim.Optimizer, pixel_criterion: nn.Module,

train.py:122

Method__call__

(self, fake_im: torch.Tensor, real_im: torch.Tensor)

(self, dataset_dir: str, mode: str)

datasetsss.py:36

Method__init__

(self, n_frames: int)

dihuman_run.py:94

Method__init__

(self, data_path: str)

dihuman_run.py:109

Method__init__

( self, inp: int, oup: int, stride: int, use_res_connect: bool,

unet.py:22

Method__init__

(self, in_channels: int, out_channels: int, stride: int = 2)

unet.py:52

Method__init__

(self, in_channels: int, out_channels: int)

unet.py:66

Method__init__

(self, in_channels: int, out_channels: int)

unet.py:76

Method__init__

(self, in_channels: int, out_channels: int)

unet.py:104

Method__init__

(self, ch=_MAIN_CHANNELS)

unet.py:115

Method__init__

(self, ch=_MAIN_CHANNELS)

unet.py:146

Method__init__

(self, n_channels: int = 6, mode: str = "wenet")

unet.py:190

Method__init__

(self, criterion: nn.Module, device: torch.device)

train.py:53

Method__init__

(self, n_frames: int)

inference.py:88

Method__init__

(self, model_config)

data_utils/wenet_infer.py:17

Method__init__

(self, engine_config)

data_utils/FeaturePipeline.py:8

Method__init__

(self, in_channel, out_channel, stride, use_res_connect, expand_ratio)

data_utils/base_module.py:21

Method__init__

(self, in_channel, out_channel, is_linear=False)

data_utils/base_module.py:44

Method__init__

(self, in_channel, hidden_channel, out_channel, stride)

data_utils/base_module.py:61

Method__init__

(self, in_channel, out_channel, is_linear=False, inference_mode=False, num_conv_branches=1)

data_utils/base_module.py:88

Method__init__

(self, in_channel, hidden_channel, out_channel, stride, inference_mode=False, num_conv_branches=1)

data_utils/base_module.py:125

Method__init__

Construct a MobileOneBlock module. :param in_channels: Number of channels in the input. :param out_channels: Number of channels prod

data_utils/base_module.py:203

Method__init__

(self, onnxmodel, confThreshold=0.5, nmsThreshold=0.5)

data_utils/detect_face.py:7

Method__init__

(self, width_factor=0.5, input_size=192, landmark_number=110, inference_mode=False)

data_utils/pfld_mobileone.py:13

Method__init__

(self, width_factor=0.5, input_size=112, landmark_number=110, inference_mode=False)

data_utils/pfld_mobileone.py:137

Method__init__

(self)

data_utils/get_landmark.py:71

Method__init__

(self)

data_utils/wenet/utils/executor.py:13

Method__init__

( self, optimizer: torch.optim.Optimizer, warmup_steps: Union[int, float] = 25000,

data_utils/wenet/utils/scheduler.py:26

Method__init__

Construct a PositionwiseFeedForward object.

data_utils/wenet/transformer/positionwise_feed_forward.py:23

Method__init__

( self, vocab_size: int, encoder_output_size: int, attention_heads: int = 4,

data_utils/wenet/transformer/decoder.py:196

Method__init__

Construct an DecoderLayer object.

data_utils/wenet/transformer/decoder_layer.py:33

Method__init__

( self, vocab_size: int, encoder: TransformerEncoder, decoder: TransformerDeco

data_utils/wenet/transformer/asr_model.py:39

Method__init__

( self, vocab_size: int, encoder_output_size: int, attention_heads: int = 4,

data_utils/wenet/transformer/decoder_streaming.py:212

Method__init__

Construct an linear object.

data_utils/wenet/transformer/subsampling.py:32

Method__init__

Construct an Conv2dSubsampling4 object.

data_utils/wenet/transformer/subsampling.py:78

Method__init__

Construct an Conv2dSubsampling6 object.

data_utils/wenet/transformer/subsampling.py:133

Method__init__

Construct an Conv2dSubsampling8 object.

data_utils/wenet/transformer/subsampling.py:185

Method__init__

( self, vocab_size: int, encoder: TransformerEncoder, decoder: TransformerDeco

data_utils/wenet/transformer/asr_model_streaming.py:39

Method__init__

Construct an LabelSmoothingLoss object.

data_utils/wenet/transformer/label_smoothing_loss.py:44

Method__init__

Construct CTC module Args: odim: dimension of outputs encoder_output_size: number of encoder projection units

data_utils/wenet/transformer/ctc.py:8

Method__init__

Args: mean (torch.Tensor): mean stats istd (torch.Tensor): inverse std, std which is 1.0 / std

data_utils/wenet/transformer/cmvn.py:20

Method__init__

Construct an RelPositionMultiHeadedAttention object.

data_utils/wenet/transformer/attention.py:144

Method__init__

Construct an ConvolutionModule object. Args: channels (int): The number of channels of conv layers. kernel_size (int):

data_utils/wenet/transformer/convolution.py:17

Method__init__

Construct TransformerEncoder See Encoder for the meaning of each parameter.

data_utils/wenet/transformer/encoder_streaming.py:345

Method__init__

Construct ConformerEncoder Args: input_size to use_dynamic_chunk, see in BaseEncoder positionwise_conv_kernel_size (i

data_utils/wenet/transformer/encoder_streaming.py:388

Method__init__

Construct TransformerEncoder See Encoder for the meaning of each parameter.

data_utils/wenet/transformer/encoder.py:319

Method__init__

Construct ConformerEncoder Args: input_size to use_dynamic_chunk, see in BaseEncoder positionwise_conv_kernel_size (i

data_utils/wenet/transformer/encoder.py:362

Method__init__

Initialize class.

data_utils/wenet/transformer/embedding.py:93

← previous next →101–200 of 297, ranked by callers