iman 1.0.22

Python package for daily Tasks

from iman import *

1-plt

2-now() get time

3-F format floating point

4-D format int number

5-Write_List(MyList,Filename)

6-Write_Dic(MyDic,Filename)

7-Read(Filename) read txt file

8-Read_Lines(Filename) read txt file line by line and return list

9-Write(_str,Filename)

10-gf(pattern) Get files in a directory

11-gfa(directory_pattern , ext=”.”) Get Files in a Directory and SubDirectories

12-ReadE(Filename) Read Excel files

13-PM(dir) creat directory

14-PB(fname) get basename

15-PN(fname) get file name

16-PE(fname) get ext

17-PD(fname) get directory

18-PS(fname) get size

19-PJ(segments) Join Path

20-clear() clear cmd

21-os

22-np

23-RI(start_int , end_int , count=1) random int

24-RF(start_float , end_float , count=1) random float

25-RS(Arr) shuffle

26-LJ(job_file_name)

27-SJ(value , job_file_name)

28-LN(np_file_name)

29-SN(arr , np_file_name)

30-cmd(command , redirect=True) Run command in CMD

31-PX(fname) check existance of file

32-RC(Arr , size=1) Random Choice

33-onehot(data, nb_classes)

34-exe(pyfile) need pyinstaller

35-FWL(wavfolder , sr) Get Folder Audio Length

36-norm(vector) vector/magnitude(vector)

37-delete(pattern)

38-rename(fname , fout)

39-separate(pattern,folout) separate vocal from music

40-dll(fname) create a pyd file from py file

41-get_hard_serial()

42-mute_mic() on and off microphone

from iman import Audio

1-Read(filename,sr,start_from,dur,mono,ffmpeg_path,ffprobe_path) Read wav alaw and mp3 and others

2-Resample(data , fs, sr)

3-Write(filename, data ,fs)

4-frame(y)

5-split(y)

6-ReadT(filename, sr , mono=True) Read and resample wav file with torchaudio

7-VAD(y,top_db=40, frame_length=200, hop_length=80)

8-compress(fname_pattern , sr=16000 , ext=’mp3’ , mono=True ,ffmpeg_path=’c:\ffmpeg.exe’ , ofolder=None, worker=4)

9-clip_value(wav) return clipping percentage in audio file

10-WriteS(filename, data ,fs) Convert to Sterio

from iman import info

1-get() info about cpu and gpu need torch

2-cpu() get cpu percentage usage

3-gpu() get gpu memory usage

4-memory() get ram usage GB

5-plot(fname=”log.txt” , delay=1)

from iman import metrics

1-EER(lab,score)

2-cosine_distance(v1,v2)

3-roc(lab,score)

4-wer(ref, hyp)

5-cer(ref, hyp)

6-wer_list(ref_list , hyp_list)

7-cer_list(ref_list , hyp_list)

8-DER(ref_list , res_list , file_dur=-1 , sr=8000) Detection Error Rate

from iman import tsne

1-plot(fea , label)

from iman import xvector

1-xvec,lda_xvec,gender = get(filename , model(model_path , model_name , model_speaker_num))

from iman import web

1-change_wallpaper()

2-dl(url)

3-links(url , filter_text=None)

4-imgs(url , filter_text=None)

from iman import matlab

1-np2mat(param , mat_file_name)

2-dic2mat(param , mat_file_name)

3-mat2dic (mat_file_name)

from iman import Features

1- mfcc_fea,mspec,log_energy = mfcc.SB.Get(wav,sample_rate) Compute MFCC with speechbrain - input must read with torchaudio

2-mfcc.SB.Normal(MFCC) Mean Var Normalization Utt with speechbrain

3- mfcc_fea,log_energy = mfcc.LS.Get(wav,sample_rate,le=False) Compute MFCC with Librosa - input is numpy array

4-mfcc.LS.Normal(MFCC , win_len=150) Mean Var Normalization Local 150 left and 150 right

from iman import AUG

1-Add_Noise(data , noise , snr)

2-Add_Reverb( data , rir)

3-Add_NoiseT(data , noise , snr) (torchaudio)

4-Add_ReverbT( data , rir) (torchaudio)

5-mp3(fname , fout,sr_out,ratio,ffmpeg_path=’c:\ffmpeg.exe’)

6-speed(fname,fout,ratio,ffmpeg_path=’c:\ffmpeg.exe’)

7-volume(fname ,fout,ratio,ffmpeg_path=’c:\ffmpeg.exe’)

from iman.[sad_torch_mfcc | sad_tf] import *

seg = Segmenter(batch_size, vad_type=[‘sad’|’vad’] , sr=8000 , model_path=”c:\sad_model_pytorch.pth” , tq=1,ffmpeg_path=’c:\ffmpeg.exe’,complete_output=False , device=’cuda’,input_type=’file’) TORCH

seg = Segmenter(batch_size, vad_type=[‘sad’|’vad’] , sr=16000 , model_path=”c:\keras_speech_music_noise_cnn.hdf5”,gender_path=”c:\keras_male_female_cnn.hdf5”,ffmpeg_path=’c:\ffmpeg.exe’,detect_gender=False,complete_output=False,device=’cuda’,input_type=’file’) TensorFlow

isig,wav,mfcc = seg(fname) mfcc output Just in torch model

nmfcc = filter_fea(isig , mfcc , sr , max_time) Just in torch model

mfcc = MVN(mfcc) Just in torch model

isig = filter_output(isig , max_silence ,ignore_small_speech_segments , max_speech_len ,split_speech_bigger_than) Do when complete_output=False

seg2aud(isig , filename)

seg2json(isig)

seg2Gender_Info(isig)

seg2Info(isig)

wav_speech , wav_noise = filter_sig(isig , wav , sr) Get Speech and Noise Parts of file - Do when complete_output=False

from sad_tf.segmentero import Segmenter to use onnx models - need to install onnxruntime

from iman.sad_torch_mfcc_speaker import *

seg = Segmenter(batch_size, vad_type=[‘sad’|’vad’] , sr=8000 , model_path=”c:\sad_model_pytorch.pth” , max_time=120(sec) , tq=1,ffmpeg_path=’c:\ffmpeg.exe’, device=’cuda’ , pad=False) TORCH - max_time in second to split fea output mfcc, len(sec) = seg(fname) mfcc pad to max_time length if pad=True

from iman.sad_tf_mlp_speaker import *

seg = Segmenter(batch_size, vad_type=[‘sad’|’vad’] , sr=8000 , model_path=”sad_tf_mlp.h5” , max_time=120(sec) , tq=1,ffmpeg_path=’c:\ffmpeg.exe’, device=’cuda’ , pad=False) Tensorflow (small mlp model) - max_time in second to split fea output mfcc, len(sec) = seg(fname) mfcc pad to max_time length if pad=True

from iman import Report Tensorboard Writer

r=Report.rep(log_dir=None)

r.WS(_type , _name , value , itr) Add_scalar

r.WT(_type , _name , _str , itr) Add_text

r.WG(pytorch_model , example_input) Add_graph

r.WI(_type , _name , images , itr) Add_image

from iman import par

if (__name__ == ‘__main__’):

res = par.par(files , func , worker=4 , args=[]) def func(fname , _args): ...

from iman import Image

Image.convert(fname_pattern ,ext =’jpg’,ofolder=None , w=-1 , h=-1,level=100, worker=4,ffmpeg_path=’c:\ffmpeg.exe’)

Image.resize(fname_pattern ,ext =’jpg’,ofolder=None , w=2 , h=2, worker=4,ffmpeg_path=’c:\ffmpeg.exe’) resize to 1/h and 1/w

from iman import Boors

Boors.get(sahm) get sahm info

from iman import Text

norm = Text.normal(“c:\Replace_List.txt”)

norm.rep(str)

norm.from_file(filename ,file_out=None)

from iman.num2fa import words

words(number)

from iman import examples

examples.items get items in examples folder

examples.help(topic)

from iman import Rar

1-rar(fname , out=”” , rar_path=r”C:\Program Files\WinRAR\winrar.exe”)

2-zip(fname , out=”” , rar_path=r”C:\Program Files\WinRAR\winrar.exe”)

3-unrar(fname , out=”” , rar_path=r”C:\Program Files\WinRAR\winrar.exe”)

4-unzip(fname , out=”” , rar_path=r”C:\Program Files\WinRAR\winrar.exe”)

from iman import Enhance

1-Enhance.Dereverb(pattern , out_fol , sr = 16000, batchsize=16 , device=”cuda” ,model_path=r”C:\UVR-DeEcho-DeReverb.pth”)

2-Enhance.Denoise(pattern , out_fol , sr = 16000, batchsize=16 , device=”cuda” ,model_path=r”C:UVR-DeNoise-Lite.pth”)

from iman.tf import *

1-flops(model) get flops of tf model

2-param(model) return parameter number of tf model

3-paramp(model) return parameter number of tf model and print model layers

4-gpu() return True if available

5-gpun() return number of gpus

6-limit() Tf model only allocate as much GPU memory based on runtime allocations

from iman.torch import *

1-param(model) return parameter number and trainable number of torch model

2-paramp(model) return parameter number of torch model and print model layers

3-layers(model) return layers of torch model

4-gpu() return True if available

5-gpun() return number of gpus