“2024-02-19”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
 
(6位用户的7个中间修订版本未显示)
第37行: 第37行:
 
** Re-write paper
 
** Re-write paper
 
** training & testing
 
** training & testing
** [https://z1et6d3xtb.feishu.cn/wiki/FinVw2p1hiPR44kBoYXcYWPRnbw work]
+
* [https://z1et6d3xtb.feishu.cn/wiki/FinVw2p1hiPR44kBoYXcYWPRnbw work]
 
||
 
||
 
*  
 
*  
第58行: 第58行:
 
|Junming Yuan
 
|Junming Yuan
 
||  
 
||  
*
+
* mix-training pretraining code checking[https://z1et6d3xtb.feishu.cn/docx/Tz4RdhYchouSGzxYRvIc2ow3nmb]
 
||
 
||
 
*
 
*
第82行: 第82行:
 
|Xiaolou Li
 
|Xiaolou Li
 
||  
 
||  
*  
+
* DeepFake test on LAV-DF
 +
* Noise test on two dataset
 
||
 
||
*  
+
* DeepFake test on FF+
 
||
 
||
 
*   
 
*   
第119行: 第120行:
 
|Wan Lin
 
|Wan Lin
 
||  
 
||  
*  
+
* Neural scoring [https://z1et6d3xtb.feishu.cn/docx/RTGSd4T1PonoF5xMajsc3OAbndh?from=from_copylink]
 
||
 
||
 
*
 
*
第145行: 第146行:
 
* Extensive Speaker Augmentation
 
* Extensive Speaker Augmentation
 
** finetune the flexible parameter of  VTLP augmentation
 
** finetune the flexible parameter of  VTLP augmentation
 +
** baseline:2.45%  vtlp aug:2.13%
 +
** warp:0.95 and 1.05, Generate training data three times the original data
 
||
 
||
 
*
 
*
第195行: 第198行:
 
||  
 
||  
 
* Write the acceptance report for the Diting project
 
* Write the acceptance report for the Diting project
 +
* update model(cn data aug)for aibabel
 
||
 
||
 
*
 
*
第215行: 第219行:
 
|Lily
 
|Lily
 
||
 
||
*  
+
* Data annotation
 +
* Interspeech2024
 
||
 
||
 
*
 
*

2024年2月19日 (一) 11:46的最后版本

People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • CH-GHUY paper half
  • ICME paper review
Lantian Li
  • GPU status [1]
  • ASIP-BUPT (Cohort-ranking SE)
  • AI course for Primary School
Ying Shi
  • Overlap ASR
    • Cohort Overlap ASR 3-Mix(20.10%)
    • CTC rank Overlap ASR 3-Mix(21.09%)
  • Phrase guided Target content extraction
    • Re-write paper
    • training & testing
  • work
Zhenghai You
  • Cohort based on Speakerbeam structure[2]
  • Cohort with Improved structure[3]
Junming Yuan
  • mix-training pretraining code checking[4]
Chen Chen
  • support child_record website
  • train multi-speaker vts model
  • read paper about deep-fake detection
Xiaolou Li
  • DeepFake test on LAV-DF
  • Noise test on two dataset
  • DeepFake test on FF+
Zehua Liu
  • DeepFake test on FF+[5]
Pengqi Li
  • Extended Paper of “phonemes contribution in SR”
    • More detailed classification of phonemes.(Consider Bi-phone)
    • Different languages.
    • text-dependent to text-independent
  • XAI review of speech field
Wan Lin
  • Neural scoring [6]
Tianhao Wang
  • reorganize SE Adapter paper
    • explanation part
    • SE weight for different genre experiments
Zhenyu Zhou
  • Extensive Speaker Augmentation
    • finetune the flexible parameter of VTLP augmentation
    • baseline:2.45% vtlp aug:2.13%
    • warp:0.95 and 1.05, Generate training data three times the original data
Junhui Chen
  • Neural Scoring
    • top-k frame score[7]
Jiaying Wang
  • pit baseline (tasnet & Convtasnet)
    • some bug(to be fixed in 2-3 days)
  • proposal of tse & ss
  • huawei project on speakerbeam
  • separation with cohort based on speech separation frame
Yu Zhang
  • portfolio analysis metric
    • AccNetValue CumReturns UnitNetValue
  • finish portfolio analysis logic
  • Investigate how to use AutoML to perform factor analysis
Wenqiang Du
  • Write the acceptance report for the Diting project
  • update model(cn data aug)for aibabel
Yang Wei
  • Fix bug of my corpus backup script
Lily
  • Data annotation
  • Interspeech2024