2024-06-24

来自cslt Wiki
2024年6月24日 (一) 11:30Lily讨论 | 贡献的版本

(差异) ←上一版本 | 最后版本 (差异) | 下一版本→ (差异)
跳转至: 导航搜索
People This Week Next Week Task Tracking (DeadLine)
Dong Wang
  • AI Primary School Plan (refinement)
  • Demo chapters of the AI book
  • Some paper review
Lantian Li
  • GPU status [1]
  • Projects
    • AED -> handover to Xiaoxue
    • TSE -> preparing for the 1st phase delivery
    • VSR -> finish the 1st phase - 1278h
    • Finance -> reproducing R^2 SAC (debug)
  • Papers
    • NeuralScoring -> in progress (multi-domain testing)
  • ASIP-Lab website -> Xiaohuang
Ying Shi
  • text enroll keyword spotting
    • Acc: 75% / FA 150/50000
  • Cohort conditional chain Overlap ASR
    • fix condition positional embedding
Zhenghai You
  • paper reading
  • preparing for the 1st phase delivery[2]
Junming Yuan
  • SSL model finetuning analysis v2[3]
    • The result is quite strange, I'll check it again
  • prepare the talk on last saturday's activity.
Chen Chen
  • Finish the draft of ISCSLP paper[4]
  • Handing over the data collection website
  • Write some documents about the websites tools
  • Refine the ISCSLP paper
Xiaolou Li
  • Experiment about bi-Transformer[5]
  • Paper Reading (Some about Mamba, Some about MLLM)
Zehua Liu
  • Some exp(current SOTA cer: 46.06%)[6]
  • Working on VSP-LLM
Pengqi Li
  • PhD mid-term assessment(pass)
  • two NC-papers(done)
Wan Lin
  • NS for CNC (single-speaker)
    • All utts as enroll
      (can't be trained directly, must load vox checkpoint)
    • Choose some genres as enroll (trying)
  • trials testing
  • Revised paper
Tianhao Wang
  • Neural Scoring [7]:
    • Soving the test set problem with CJH
    • Try NS on CN or "Vox mix by CN"
Zhenyu Zhou
  • preparing for the 1st phase delivery[8]
Junhui Chen
  • Neural Scoring:
    • Vox-E/H trials testing
    • Try test utt: Vox mix CN [9]
Jiaying Wang
  • water project....
Yu Zhang
  • Reproduce R^2-SAC
    • Code is complete
    • Adapting training and testing process to our backtesting platform (still working)
  • Get more data and train the whole process align with Jun Wang's paper (2015.01-2023.6 CSI300 CSI500 is needed)
Wenqiang Du
  • Training dialect models for AIbabel(minnan language)
Yang Wei
  • Huilan
    • Some support for ASR project
    • Fix bugs on TTS code about ONNX model export
Lily
  • Organize AI Graph material (动动脑筋)
  • Organize primary, high school AI courses materials
Turi
  • Thesis Proposal
  • Data collection
    • 28.9K contributions so far
Yue Gu
Qi Qu
  • KWD:
    • More lab-indoor data collected.
  • AED:
    • AED model (frontend) + Linear (classifier).
    • CED (frontend) models yield acceptable results.
  • AED:
    • CED + Linear: c/jni lib.