“2024-05-13”版本间的差异

来自cslt Wiki
跳转至： 导航、 搜索
 
（19位用户的24个中间修订版本未显示）
第6行： 第6行：
 
|Dong Wang
 
|Dong Wang
 
||  
 
||  
−
*  
+
 
 +
* Material preparation for Xinhua Net broadcast
 +
* Several public reports
 +
* Review for Electonics and Applied Science
 
||
 
||
 
*
 
*
第17行： 第20行：
 
|Lantian Li
 
|Lantian Li
 
||
 
||
−
*  
+
* GPU status [https://z1et6d3xtb.feishu.cn/wiki/XGcGwRK5viJmpRkjH9AczIhynCh]
 +
* Projects (AED -> Hardware support, TSE -> Test&Analysis)
 +
* ASIP-BUPT (NeuralScoring -> Paper, CohortSS -> Data Analysis)
 +
* Check NIPS & Review theses
 
||
 
||
 
*  
 
*  
第23行： 第29行：
 
*   
 
*   
 
|-
 
|-
 +
  
  
第28行： 第35行：
 
|Ying Shi
 
|Ying Shi
 
||  
 
||  
−
*
+
* verify cohort Overlap ASR assumption
 +
** Identify the speech component which most similar to the cohort vector ✔
 +
* [https://z1et6d3xtb.feishu.cn/docx/PBaLdj17ao7mKaxFNPacr3aYn8c?from=from_copylink group work]
 
||
 
||
−
*  
+
* cohort + conditional chain Overlap ASR
 
||
 
||
 
*   
 
*   
第39行： 第48行：
 
|Zhenghai You
 
|Zhenghai You
 
||  
 
||  
−
*  
+
* Speech tests and deliver real test samples for HUAWEI
 +
* Loudness testing and adjustment of Huawei data[https://z1et6d3xtb.feishu.cn/docx/SFZBdrHafohmQJx1ti7c2RZwnuf]
 +
* Comparative experiments on data expansion
 
||
 
||
 
*  
 
*  
第49行： 第60行：
 
|Junming Yuan
 
|Junming Yuan
 
||  
 
||  
−
*
+
* Continue to add various data augmentation functions into the code
 +
* Prepare for live broadcast
 
||
 
||
 
*
 
*
第60行： 第72行：
 
|Chen Chen
 
|Chen Chen
 
||  
 
||  
−
*  
+
* attend several interviews for job
 +
* vii group work [https://z1et6d3xtb.feishu.cn/docx/GwFvdn3nnopuU4xhKUncTxSnnTg?from=from_copylink]
 
||
 
||
 
*  
 
*  
第71行： 第84行：
 
|Xiaolou Li
 
|Xiaolou Li
 
||  
 
||  
−
*  
+
* Video mamba exp (good good)
 +
** patch frontend
 +
** conv3d and resnet3d frontend
 +
* Paper reading
 
||
 
||
−
*  
+
* run exp on LRS2 and LRS3 (waiting for email feedback)
 +
* what is the main difference between these two frontend? (conv3d and resnet3d)
 
||
 
||
 
*   
 
*   
第82行： 第99行：
 
|Zehua Liu
 
|Zehua Liu
 
||  
 
||  
−
*
+
*AKVSR (cer:49.71%) > baseline(cer: 48.76%)
 +
**AKVSR + pos_emb (a little worse)
 +
**AKVSR + attention score loss(coding)
 
||
 
||
 
*  
 
*  
第93行： 第112行：
 
|Pengqi Li
 
|Pengqi Li
 
||   
 
||   
−
*  
+
* Jinfu and LiuHuan's Outlines of NC
 
||
 
||
−
*
+
* XueYing's Outline of NC
 +
* NC paper of Speech XAI overview
 
||
 
||
 
*   
 
*   
第104行： 第124行：
 
|Wan Lin
 
|Wan Lin
 
||  
 
||  
−
*  
+
* EAASP in Sunine(EER)
 +
** EA:4.292(3.106 wespeaker）
 +
** Mix: 7.733(5.962 wespeaker）
 +
* Add CNN condition in test encoder: currently unsuccessful
 
||
 
||
 
*
 
*
第115行： 第138行：
 
|Tianhao Wang
 
|Tianhao Wang
 
||  
 
||  
−
*  
+
* Baseline: SpEx+ with Detection (Failed)
 +
** difficult to train because vox2 has a much larger data volume than wsj0
 +
* Toolkit align: lr scheduler, pooling
 +
** pooling seems critical (same epoch, NS loss: ASP is 0.16 vs TSP is 0.22)
 
||
 
||
 
*  
 
*  
第126行： 第152行：
 
|Zhenyu Zhou
 
|Zhenyu Zhou
 
||  
 
||  
−
*
+
*HUAWEI project process[https://z1et6d3xtb.feishu.cn/docx/PBAZdsiSWoq82YxWsu3cCD4Tnte]
 
||
 
||
 
*
 
*
第137行： 第163行：
 
|Junhui Chen
 
|Junhui Chen
 
||
 
||
−
*  
+
* Graduation paper
 +
* Neural Scoring paper writing
 
||
 
||
 
*
 
*
第148行： 第175行：
 
|Jiaying Wang
 
|Jiaying Wang
 
||  
 
||  
−
*  
+
* find bad cases in the test set(gender confusion)
 
||
 
||
−
*  
+
* data analyse
 +
* focus on cohort outside masker
 
||
 
||
 
*   
 
*   
第159行： 第187行：
 
|Yu Zhang
 
|Yu Zhang
 
||
 
||
−
*  
+
* AutoML
 +
** EvalML test result[https://z1et6d3xtb.feishu.cn/docx/EDO1dLwHToDqiCxhHf6cLXDVnlb?from=from_copylink]
 
||
 
||
 
*
 
*
第170行： 第199行：
 
|Wenqiang Du
 
|Wenqiang Du
 
||  
 
||  
−
*  
+
* Just some project test
 
||
 
||
 
*
 
*
第181行： 第210行：
 
|Yang Wei
 
|Yang Wei
 
||  
 
||  
−
*  
+
* Children MDD challenge
 +
** Refine documentation and prepare material for discuss
 +
* Huilan stuff
 +
** Reduce size of TTS Docker image
 
||
 
||
 
*
 
*
第191行： 第223行：
 
|Lily
 
|Lily
 
||
 
||
−
* PPT delivery
+
* AIGraph PPT delivery
 
* Thesis  
 
* Thesis  
−
* Perception experiment
+
* Perception Experiment
 
||
 
||
 
*
 
*
第213行： 第245行：
 
|Yue Gu
 
|Yue Gu
 
||
 
||
−
*  
+
* fail to reproduct the semantic paraformer
 +
* write paper：30% of experimental part
 +
* kespeech baseline
 
||
 
||
 
*
 
*
第222行： 第256行：
 
|Qi Qu
 
|Qi Qu
 
||
 
||
−
*  
+
* KWS
 +
** Standardize dataset formats and test routines.
 +
** Data collection and processing.
 
||
 
||
 
*
 
*

2024年5月13日 (一) 11:22的最后版本

People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • Material preparation for Xinhua Net broadcast
  • Several public reports
  • Review for Electonics and Applied Science
Lantian Li
  • GPU status [1]
  • Projects (AED -> Hardware support, TSE -> Test&Analysis)
  • ASIP-BUPT (NeuralScoring -> Paper, CohortSS -> Data Analysis)
  • Check NIPS & Review theses
Ying Shi
  • verify cohort Overlap ASR assumption
    • Identify the speech component which most similar to the cohort vector ✔
  • group work
  • cohort + conditional chain Overlap ASR
Zhenghai You
  • Speech tests and deliver real test samples for HUAWEI
  • Loudness testing and adjustment of Huawei data[2]
  • Comparative experiments on data expansion
Junming Yuan
  • Continue to add various data augmentation functions into the code
  • Prepare for live broadcast
Chen Chen
  • attend several interviews for job
  • vii group work [3]
Xiaolou Li
  • Video mamba exp (good good)
    • patch frontend
    • conv3d and resnet3d frontend
  • Paper reading
  • run exp on LRS2 and LRS3 (waiting for email feedback)
  • what is the main difference between these two frontend? (conv3d and resnet3d)
Zehua Liu
  • AKVSR (cer:49.71%) > baseline(cer: 48.76%)
    • AKVSR + pos_emb (a little worse)
    • AKVSR + attention score loss(coding)
Pengqi Li
  • Jinfu and LiuHuan's Outlines of NC
  • XueYing's Outline of NC
  • NC paper of Speech XAI overview
Wan Lin
  • EAASP in Sunine(EER)
    • EA:4.292(3.106 wespeaker）
    • Mix: 7.733(5.962 wespeaker）
  • Add CNN condition in test encoder: currently unsuccessful
Tianhao Wang
  • Baseline: SpEx+ with Detection (Failed)
    • difficult to train because vox2 has a much larger data volume than wsj0
  • Toolkit align: lr scheduler, pooling
    • pooling seems critical (same epoch, NS loss: ASP is 0.16 vs TSP is 0.22)
Zhenyu Zhou
  • HUAWEI project process[4]
Junhui Chen
  • Graduation paper
  • Neural Scoring paper writing
Jiaying Wang
  • find bad cases in the test set(gender confusion)
  • data analyse
  • focus on cohort outside masker
Yu Zhang
  • AutoML
    • EvalML test result[5]
Wenqiang Du
  • Just some project test
Yang Wei
  • Children MDD challenge
    • Refine documentation and prepare material for discuss
  • Huilan stuff
    • Reduce size of TTS Docker image
Lily
  • AIGraph PPT delivery
  • Thesis
  • Perception Experiment
Turi
  • Data Collection
    • Checking audios
  • Class works
Yue Gu
  • fail to reproduct the semantic paraformer
  • write paper：30% of experimental part
  • kespeech baseline
Qi Qu
  • KWS
    • Standardize dataset formats and test routines.
    • Data collection and processing.
取自“http://index.cslt.org/mediawiki/index.php?title=2024-05-13&oldid=41767”