“2021-11-01”版本间的差异

来自cslt Wiki
跳转至: 导航搜索
第94行: 第94行:
 
|Weida Liang
 
|Weida Liang
 
||   
 
||   
*  
+
* several models tested including model without cycle loss, models with different bottleneck
 
||
 
||
*  
+
* finetune these models decoders
 
||
 
||
 
*   
 
*   

2021年11月1日 (一) 01:35的版本

People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • Anti-spoof chapter
  • Anti-spoof chapter
Yunqi Cai
  • THS dataset experiments and paper
Lantian Li
  • Submit CNC response
  • Complete project deliveries
  • Hard trials
Ying Shi
  • Continue on KWS
Haoran Sun
Chen Chen
Pengqi Li
  • egs of thc30 base sunine
  • reproduce bubble noise on speech recognition
  • audio important maps based bubble noise on speaker recognition
Qingyang Zhu
Weida Liang
  • several models tested including model without cycle loss, models with different bottleneck
  • finetune these models decoders
Zixi Yan
  • analyse reason of the problem of wav2vec-u in Tibetan
Sirui Li
  • Gan network training code
  • Mongolian recording project
  • Adjust Tibetan wav2vec-u parameters
Haoyu Jiang
  • Voxceleb2 and Voxceleb1 data set collation
  • single mode sota system training
Ruihai Hou