Python package for daily Tasks
Project description
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
from iman import Audio
1-Read(filename,sr,ffmpeg_path) Read wav alaw and mp3 (return Just MONO)
2-Resample(data , fs, sr)
3-Read_Alaw(filename)
4-ReadMp3(filename,sr,mono,ffmpeg_path) Just Windows
5-Write(filename, data ,fs)
6-frame(y)
7-split(y)
8-ReadT(filename, sr , mono=True) Read and resample wav file with torchaudio
9-VAD(y,top_db=40, frame_length=200, hop_length=80)
10-ReadMp3_miniaudio (filename,sr,mono)
11-compress(fname , sr=16000 , ext=’mp3’ , mono=True ,ffmpeg_path=’c:\ffmpeg.exe’ ,oname=None)
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)
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)
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 = mfcc.LS.Get(wav,sample_rate) 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) Don't need sox
2-Add_Reverb( data , rir) Don't need sox
3-Add_NoiseT(data , noise , snr) Don't need sox (torchaudio)
4-Add_ReverbT( data , rir) Don't need sox (torchaudio)
x=AUG.aug(sox_path) Use this Just in WINDOWS
5-x.mp3(fname , sr, fout,ratio)
6-x.speed(fname,fout,ratio)
7-x.volume(fname ,fout,ratio)
from iman.[sad_torch_mfcc | sad_tf] import *
seg = Segmenter(batch_size, vad_type=[‘sad’|’vad’] , sr=[8000 | 16000] , model_path=[“c:\sad_model_pytorch.pth” | “c:\keras_speech_music_noise_cnn.hdf5”] , max_time=120 , tq=1) max_time in second and tq(verbose) Just in torch model to split fea output
isig,wav,mfcc = seg(fname) mfcc output Just in torch model--> Concat Mfccs where speech detected
mfcc = MVN(mfcc) Just in torch model
isig = filter_output(isig , max_silence ,ignore_small_speech_segments , max_speech_len ,split_speech_bigger_than)
seg2aud(isig , filename)
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.