“2021-12-13”版本间的差异

来自cslt Wiki
跳转至： 导航、 搜索
 
（9位用户的12个中间修订版本未显示）
第5行： 第5行：
 
|Dong Wang
 
|Dong Wang
 
||  
 
||  
−
*  
+
* Spoof paper refined
 +
* Start the hard trials paper
 
||
 
||
−
*  
+
* Hard trials paper
 
||
 
||
 
*   
 
*   
第16行： 第17行：
 
|Yunqi Cai
 
|Yunqi Cai
 
||   
 
||   
−
*  
+
* img fusion network construction
 +
* infra experiments plan for interns
 +
* bayesian optimization paper review
 
||
 
||
 
*  
 
*  
第27行： 第30行：
 
|Lantian Li
 
|Lantian Li
 
||  
 
||  
−
*  
+
* Refine AI course v2.
 +
* Check spoof paper.
 +
* Finish my defences.
 
||
 
||
−
*  
+
* Finish ETM response.
 +
* Exps of hard trials.
 
||
 
||
 
*   
 
*   
第38行： 第44行：
 
|Ying Shi
 
|Ying Shi
 
||  
 
||  
−
* Test fncmd and speech engrave on huawei_cross_channel data
+
* Test fncmd and speech engrave on huawei_cross_channel data [http://cslt.riit.tsinghua.edu.cn/mediawiki/images/8/82/Speech_engrave_fncmd_huawei_cross.png here]
 
||
 
||
 
* Retrain speech engrave model(make speech engrave and fncmd are Comparable on far field test set)
 
* Retrain speech engrave model(make speech engrave and fncmd are Comparable on far field test set)
第44行： 第50行：
 
** Score margin
 
** Score margin
 
** Discriminative training
 
** Discriminative training
 +
* Retrain fncmd model with huawei data.
 
||
 
||
 
*   
 
*   
第52行： 第59行：
 
|Haoran Sun
 
|Haoran Sun
 
||  
 
||  
−
*  
+
* some analysis on c-vector
 +
* training processing of c-vector
 
||
 
||
−
*  
+
* remove f0 decoder of c-vector
 +
* a easier model with only content and speaker encoders based on long-short term assumption
 
||
 
||
 
*   
 
*   
第63行： 第72行：
 
|Chen Chen
 
|Chen Chen
 
||  
 
||  
−
*  
+
* perform kmeans and pca on wav2vec result
 +
* check GAN
 
||
 
||
−
*  
+
* fix bug of uasr_model
 
||
 
||
 
*   
 
*   
第109行： 第119行：
 
|Sirui Li
 
|Sirui Li
 
||   
 
||   
−
*  
+
* Compare the effects of TIMIT and Tibetan fine-tune
 
||
 
||
−
*  
+
* More comparative experiments
 
||
 
||
 
*   
 
*   
第120行： 第130行：
 
|Haoyu Jiang
 
|Haoyu Jiang
 
||  
 
||  
−
*  
+
* Resampling the data
 
||
 
||
−
*  
+
* Set thresholds to divide data
 +
* Check the sampled images
 
||
 
||
 
*   
 
*   
 
|-
 
|-
  
−
 
−
|-
 
−
|Ruihai Hou
 
−
||
 
−
*
 
−
||
 
−
*
 
−
||
 
−
* 
 
−
|-
 
  
 
|-
 
|-
 
|Renmiao Chen
 
|Renmiao Chen
 
||  
 
||  
−
*  
+
* choose thresholds for dividing high-confident data, mid-confident data, low-confident data.
 +
* check the thresholds.
 +
* use speechbrain to do IDR task.
 
||
 
||
−
*  
+
* do more task with the data.
 +
* finish the report.
 
||
 
||
 
*   
 
*   

2021年12月13日 (一) 11:24的最后版本

People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • Spoof paper refined
  • Start the hard trials paper
  • Hard trials paper
Yunqi Cai
  • img fusion network construction
  • infra experiments plan for interns
  • bayesian optimization paper review
Lantian Li
  • Refine AI course v2.
  • Check spoof paper.
  • Finish my defences.
  • Finish ETM response.
  • Exps of hard trials.
Ying Shi
  • Test fncmd and speech engrave on huawei_cross_channel data here
  • Retrain speech engrave model(make speech engrave and fncmd are Comparable on far field test set)
    • Huawei cross channel data
    • Score margin
    • Discriminative training
  • Retrain fncmd model with huawei data.
Haoran Sun
  • some analysis on c-vector
  • training processing of c-vector
  • remove f0 decoder of c-vector
  • a easier model with only content and speaker encoders based on long-short term assumption
Chen Chen
  • perform kmeans and pca on wav2vec result
  • check GAN
  • fix bug of uasr_model
Pengqi Li
  • Verifying the correctness of the a series of cam method
  • reproduce the method of Layer-CAM on classification
  • more experiment and analysis on this method
Weida Liang
  • Finish training for not-ever-seen speaker on baseline AE and cycle model
  • Build the framework of wav2vec model
  • Full test on baseline & cycle model
  • More details need to be discussed on wav2vec model
Zixi Yan
  • Fine-tune the wav2vec model on dev-other
  • Test the effect of Tibetan adjusted model
Sirui Li
  • Compare the effects of TIMIT and Tibetan fine-tune
  • More comparative experiments
Haoyu Jiang
  • Resampling the data
  • Set thresholds to divide data
  • Check the sampled images
Renmiao Chen
  • choose thresholds for dividing high-confident data, mid-confident data, low-confident data.
  • check the thresholds.
  • use speechbrain to do IDR task.
  • do more task with the data.
  • finish the report.
取自“http://index.cslt.org/mediawiki/index.php?title=2021-12-13&oldid=37435”