VoxSRC20
Record my story on VoxSRC20
ShaneRun
To Think, To Explore, To Contribute.
展开
-
Summary on VoxSRC20
My simple storyOne laymen and hobbyist, start from zeroLearn and use deep learningJoin the first challenge on speaker recognition as practiceStatus/TaskStatus: Use part-time to learn and practice.Task: Join VoxSRC20 and make it to TOP10ActionU原创 2020-11-04 12:09:01 · 101 阅读 · 0 评论 -
[voxsrc20_ask_27] How to get DET plot for a speaker recognition system?
ID = voxsrc20_ask_27Status: closedQuestionHow to get DET plot for a speaker recognition system?DET plot: Detection Error Trade-off plot[1]AnswerUse python to make it, see [2].Notes: Tested okay.import matplotlib.pyplot as pltfrom sklearn.metrics原创 2020-09-16 12:44:07 · 169 阅读 · 0 评论 -
voxsrc20_tsk_03-用ZeroTier搭建虚拟局域网实现内网穿透
ID = voxsrc20_tsk_03Status: closed文章目录精华TaskExecution[200816] First try1. Use requirement2. Solution3. How to?Step1: Register account of zerotier and create your NetworkStep2: install zerotier in serverStep3: install zerotier in notebookStep4: join to ta原创 2020-08-25 14:46:14 · 471 阅读 · 0 评论 -
[200724]什么才是高速固态硬盘?
更换系统盘,但是不知从何入手。。。旧系统盘:Intel SSDPEKKR256G7https://www.mouser.cn/datasheet/2/612/e6000p-product-brief-1369263.pdf看来这个就是高速固态硬盘,升级到1T,必须满足:接口:M.2接口协议:NVMe协议。下面这个网友的回答很给力,五分好评!并不是M.2接口就一bai定是高速SSD固态硬盘啦。 M.2接口标准zhi规范中,分几种规格dao,一种是走SATA总线,一种是PCIE X2总线规原创 2020-07-24 22:01:36 · 2007 阅读 · 0 评论 -
voxsrc20_ask_13-Why convert AAC to WAV?
ID = voxsrc20_ask_13Status: open文章目录QuestionAnswerReferenceQuestionWhy taking a long time to convert from aac to wav in voxceleb?$ python ./dataprep.py --save_path home/data/voxceleb --convertConverting files from AAC to WAV100%|████████████████████原创 2020-07-20 11:22:58 · 119 阅读 · 0 评论 -
voxsrc20_tsk_00-How to check md5?
How to check md5?The command is $ md5sum <target_file> | cut -d ' ' -f1Record:$ md5sum vox2_dev_aac_partaa | cut -d ' ' -f1da070494c573e5c0564b1d11c3b20577downloaded filemd5_refmd5_calcresultvox1_dev_wav_partaae395d020928bc15670b57原创 2020-07-20 09:31:40 · 173 阅读 · 0 评论 -
voxsrc20_ask_05-Understanding of PLDA
ID = voxsrc20_ask_05Status: closed文章目录QuestionAnswerExtension - Cosine distanceReferenceQuestionPLDA is widely used, how to understand it?AnswerPLDA can be treated as LDA with a probability distribution attached to the features.[1]The probability原创 2020-07-18 15:52:56 · 131 阅读 · 0 评论 -
voxsrc20_ask_01-Is embeddings the voiceprint?
ID = voxsrc20_ask_01Status: closedQuestionIs embeddings the voiceprint?Answer[200710] My first answerYES[200712] My second answerYESUsing neural network to extract the corresponding embeddings inside all utterances of each speaker.In other words,原创 2020-07-16 22:29:55 · 61 阅读 · 0 评论 -
voxsrc20_ask_00-How to understand intra-class and inter-class distance in speaker recoginition?
ID = voxsrc20_ask_00Status: closed文章目录QuestionAnswerReferenceQuestionHow to understand these two distances?Since open-set speaker recognition is essentially a metric learning problem, the key is to learn features thathave small intra-class and large原创 2020-07-16 22:23:26 · 187 阅读 · 0 评论 -
voxsrc20_ask_08-What is speaker space?
ID = voxsrc20_ask_08Status: closedContentQuestionAnswerReferenceQuestionWhat is speaker space?Is it 2D or 3D, and what is the axis?AnswerSpeaker space is high dimensional space, for example 256 or 512 dimensions.It is highly abstract space, not jus原创 2020-07-16 22:18:06 · 141 阅读 · 0 评论 -
voxsrc20_ask_07-Why there are designed to have 2 layers of embeddings in x-vector?
ID = voxsrc20_ask_07Status: closedContentQuestionAnswerReferenceQuestionWhy there are designed to have 2 layers of embeddings in x-vector?AnswerAsk1:Leon老师,您好!看了您ICASSP的论文,有一个疑问,还请解惑,谢谢!TDNN中为何设置两层segment layer?1. Pooling 之后经过segment layer1已经得到了原创 2020-07-16 21:50:08 · 98 阅读 · 0 评论 -
[200709] How to make a proposal on speaker verification?
[200709] How to make a proposal on speaker verification?categoryitemvalueacousticfeature40d mel-filter backfrontendtopologyresnet-34 + relu + BNfrontendloss functionangular prototypicalfrontendpoolingstatistic poolingfronten原创 2020-07-11 14:46:38 · 183 阅读 · 0 评论 -
voxsrc20_std_00-How many kinds of topology used in speaker recognition?
ID = voxsrc20_std_00Status: closedContentTopicStudy record[200711] VoxSRC19ReferenceTopicHow many kinds of topology used in speaker recognition?Study record[200711] VoxSRC19Mainly 2 kinds: TDNN and ResNetOverview of Top3 team’s topology:team原创 2020-07-11 14:42:03 · 106 阅读 · 0 评论 -
voxsrc20_isu_01-DL server suspend automatically
ID = voxsrc20_isu_01Status: closedServer suspend automaticallyDescriptionAnalysisActionTry to find out suspend logTry to locate which service will result in this issue?SolutionLesson learnDescriptionServer abnormally suspend after DL environment was al原创 2020-07-08 07:11:00 · 78 阅读 · 0 评论 -
voxsrc20_isu_00-Command ‘nvidia-smi‘ not found error
ID = voxsrc20_isu_00Status: closed文章目录DescriptionAnalysisAction[200702] Login the server locallySolutionLesson learnDescriptionFailed to run nvidia-smi$ nvidia-smiCommand 'nvidia-smi' not found, but can be installed with:sudo apt install nvidia-ut原创 2020-07-07 07:17:51 · 907 阅读 · 0 评论 -
voxsrc_tsk_01-How to build a deep learning server(如何搭建深度学习服务器)?
How to build a deep learning server?[200625] ProposalStep by step operationStep1: UbuntuStep2: AnacondaStep3: CUDAStep4: cuDNN[1]Step5: PytorchStep6: PycharmStep6.1: extract pycharm-community-2020.1.2Step6.2: follow `Install-Linux-tar.txt`Step6.3: add pych原创 2020-07-05 17:11:53 · 239 阅读 · 0 评论 -
voxsrc20_tsk_01-how to install pycharm and add to favorites in Ubuntu 20.04?
Install PycharmStep1: extract pycharm-community-2020.1.2Step2: follow `Install-Linux-tar.txt`Step3: add pycharm to my favoritesFirstly, download the installation package from jetbrains: pycharm-community-2020.1.2.tar.gzStep1: extract pycharm-community-20原创 2020-07-04 21:49:07 · 325 阅读 · 0 评论 -
voxsrc_tsk_02-How to control server using Remmina based on RDP (Ubuntu to Ubuntu)?
ID = voxsrc20_tsk_02文章目录TaskTargetPlanExecution[200626] Remmina or Teamviewer?[200626] Operation recordStep1: Install xrdp in serverStep2: Add to user group in serverStep3: Connect to server using RemminaReferenceFeedbackOutputTaskHow to remote control原创 2020-06-28 13:57:32 · 109 阅读 · 0 评论 -
voxsrc20_tsk_00-How to get VoxCeleb dataset?
ID = voxsrc20_tsk_00文章目录TaskExecutionSolution 1: Use official linkSolution 2: BaiduNetDisk (only VoxCeleb2)ConclusionTaskDownload VoxCeleb datasetExecutionSolution 1: Use official linkUse the official link as listed in the following table.My experie原创 2020-06-26 07:27:55 · 386 阅读 · 0 评论