|
|
| （5位用户的5个中间修订版本未显示） |
| 第71行： |
第71行： |
| | |Xiaolou Li | | |Xiaolou Li |
| | || | | || |
| − | * | + | * Prepared and attended NCMMSC 2025 |
| | || | | || |
| | * | | * |
| 第82行： |
第82行： |
| | |Zehua Liu | | |Zehua Liu |
| | || | | || |
| − | * | + | *VSR topic training for midterm（cer：39% better than 41%） |
| | || | | || |
| | * | | * |
| 第105行： |
第105行： |
| | |Wan Lin | | |Wan Lin |
| | || | | || |
| − | * | + | * NCMMSC 2025 |
| | || | | || |
| | * | | * |
| 第116行： |
第116行： |
| | |Tianhao Wang | | |Tianhao Wang |
| | || | | || |
| − | * | + | * writing paper |
| | + | * Midreport |
| | || | | || |
| | * | | * |
| 第127行： |
第128行： |
| | |Xiaoxue Luo | | |Xiaoxue Luo |
| | || | | || |
| − | * | + | * train of attractor-based speech separation |
| | + | ** The loss is abnormal, possibly due to the limited distinguishability of the mask, and further adjustments are needed |
| | + | * Preparing for my thesis proposal |
| | || | | || |
| | * | | * |
| 第165行： |
第168行： |
| | |Yu Zhang | | |Yu Zhang |
| | || | | || |
| − | * Training MIC to VPU convert for Huawei AED | + | * Training MIC to VPU converter for Huawei AED |
| | * Keeps running & debugging reflection metric collection with @chenjunhui | | * Keeps running & debugging reflection metric collection with @chenjunhui |
| | || | | || |
| 第189行： |
第192行： |
| | |Yang Wei | | |Yang Wei |
| | || | | || |
| − | * | + | * Fix some bugs of the MD training code. (Not work, bce loss > 0.4) |
| | + | ** stop inserting special token on phone sequence |
| | + | ** correct Whisper speech embedding batch mask |
| | + | ** add segment embedding and separate token embedding to identify the two modality |
| | || | | || |
| | * | | * |
| People |
This Week |
Next Week |
Task Tracking (DeadLine)
|
| Dong Wang
|
- Travel to Zhenjiang
- AI+ project design
- AI textbook 2nd semester, primary school version (3 grad)
|
|
|
| Lantian Li
|
- Review AI Book – College Edition (40/53).
- Submit Professional title evaluation.
- Prepare the 1st delivery report for HUAWEI.
|
|
|
| Ying Shi
|
- prepared for interviews
- report some experiments on MD tasks
|
|
|
| Zhenghai You
|
- AATC2025 baseline report in NC
- Support for the Huawei SS project
- Midreport
|
|
|
| Junming Yuan
|
- Evaluating the performance of MT-HuBERT on the speaker diarization task.
- DER(%)result: MT-HuBERT(3.93) < Cocktail-HuBERT(4.07) < WavLM-BASE(4.55) < MPC-HuBERT(4.83) < HuBERT-BASE(5.88)
- Start checking the resources for the middle-school AI course with Pengqi and Qiang bro(PPT, teaching plans).
|
|
|
| Xiaolou Li
|
- Prepared and attended NCMMSC 2025
|
|
|
| Zehua Liu
|
- VSR topic training for midterm（cer：39% better than 41%）
|
|
|
| Pengqi Li
|
- Try to Explain for hard-trials(Prof.Li previous work).
- Outline the paper new structure.
|
|
|
| Wan Lin
|
|
|
|
| Tianhao Wang
|
|
|
|
| Xiaoxue Luo
|
- train of attractor-based speech separation
- The loss is abnormal, possibly due to the limited distinguishability of the mask, and further adjustments are needed
- Preparing for my thesis proposal
|
|
|
| Junhui Chen
|
- Keeps running & debugging reflection metric collection with @ZhangYu
- Collecting & reading papers from NIPS & COLM 2025
|
|
|
| Jiaying Wang
|
- convtasnet pit & loudness order exp and visulization analyse
- mid-term report
- Report framework diagram
- Report progress: The first part of the research and the PIT-related experimental section have been completed.
|
|
|
| Yu Zhang
|
- Training MIC to VPU converter for Huawei AED
- Keeps running & debugging reflection metric collection with @chenjunhui
|
|
|
| Wenqiang Du
|
- BUPT anniversary exhibition preparation
- Check primary AI handbook（done）and related PPT、Jiaoan(6/44)
|
|
|
| Yang Wei
|
- Fix some bugs of the MD training code. (Not work, bce loss > 0.4)
- stop inserting special token on phone sequence
- correct Whisper speech embedding batch mask
- add segment embedding and separate token embedding to identify the two modality
|
|
|
| Yue Gu
|
- prepare for the writing test of the civil service exam
- almost finish the structure of my thesis
|
|
|
| Qi Qu
|
- Development for elderly alarm customization.
- Perf test for vision tasks (face detection, face tracking, etc.) across candidate ROS main boards.
- (In progress) Run vision tasks w/ Rockchip chips.
|
|
|
| Lily
|
- Paper reading
- Organize AI handbook English version course materials
|
|
|
