Scientific Computing (科學計算)

Roger Jang (張智星)

Final Project:

For final project, we need to participate the task of singing transcription in AI-Cup competition.

• You need to form a team of 3 or less members.
• You need to register for the task of singing transcription of AI Cup, and the deadline is 2020/06/01 (Monday).
• You also need to fill the participation table, so TA will be able to check out your performance in AI-Cup.
• For discussion about the task, please refer to Facebook for Singing Transcription in AI Cup.
• The result submission deadline is 2020/06/22 (same as the deadline of AI Cup).
• The report submission deadline for our class is 2020/06/24.
• Grading policy:
  • (50%) The ranking of your performance in the competition.
  • (50%) The quality and completeness of your report (max. 4 pages).
• Each team needs to submit a report in PDF format, up to 4 pages. (If your submission is more than 4 pages, the extra pages will not be graded.) The report should covers at least the following items (10% each):
  • The models you tried.
  • Dataset partitioning (training, validation, and test sets) and the corresponding performance evaluation (holdout test, m-fold cross validation, etc.)
  • Feature normalization, selection, and extraction
  • Improvement over the baseline method
  • Insight and conclusions
  • Division of labor (請列出組員之間如何分工).
  補充：如果你的方法和機器學習的相關度比較低，你的報告請偏重於下列幾點：說明你的想法和你的方法、如何逐步驗證你的方法的適用性、如何根據你的發現來逐步調整方法、如何調整參數、如何提高效能，等等。（這些原則也適用於採用機器學習方法的報告。）
• Videos
  • Introduction to the task of singing transcription (歌唱轉譜) in AI Cup
  • Introduction to Singing Transcription
  • Introduction to Singing Voice Separation
  • Audio Signal Processing
• Slides/videos at Scientific Computing
  • Intro to singing transcription: slides, video
  • Visualization and rendering of PV and music notes: slides, video
  • Basic methods for singing transcription: slides, video
• My MATLAB Tutorial
  • Part-1: Visualization/rendering
  • Part-2: Transcription
  • Part-3: Performance evaluation