Skip to content

Latest commit

ย 

History

History
339 lines (250 loc) ยท 18.1 KB

README.md

File metadata and controls

339 lines (250 loc) ยท 18.1 KB

maum.ai ๋ธŒ๋ ˆ์ธ ๋ธ”๋กœ๊ทธ

Table of Contents


Local์—์„œ ์‚ฌ์ดํŠธ ๋ Œ๋”๋งํ•˜๊ธฐ

requirements

  • Node.js >= 14

์„ค์น˜ํ•˜๊ธฐ

yarn install --frozen-lockfile

๋ Œ๋”๋งํ•˜๊ธฐ

Hot Reload๋ฅผ ์ง€์›ํ•˜์—ฌ ์ˆ˜์ •ํ•˜์ž๋งˆ์ž ๋ฐ”๋กœ๋ฐ”๋กœ ํ™•์ธํ•˜์‹ค ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

yarn run start

์ ‘์†ํ•˜๊ธฐ

  • ์ถœ๋ ฅ๋œ ์ฃผ์†Œ(ex. 127.0.0.1:3000)์— ์›น๋ธŒ๋ผ์šฐ์ €๋ฅผ ์‚ฌ์šฉํ•ด ์ ‘์†ํ•ฉ๋‹ˆ๋‹ค.

Branch ๊ด€๋ฆฌ

์ฃผ์š” branch

  • gh-pages: public ๊ณต๊ฐœ๋˜๋Š” branch์ž…๋‹ˆ๋‹ค. ์—…๋ฐ์ดํŠธ ๋˜๋ฉด ์‚ฌ์ดํŠธ ์ปดํŒŒ์ผ์— ํ•„์š”ํ•œ ์‹œ๊ฐ„์ด ์ง€๋‚œ ํ›„ mindslab-ai.github.io์— ๋ฐ˜์˜๋ฉ๋‹ˆ๋‹ค.
  • master: build ํ•  ์ˆ˜ ์žˆ๋Š” docusaurus template์ด ์žˆ๋Š” branch. push ์‹œ github action์— ์˜ํ•ด ์ž๋™ build ๋˜์–ด push
  • contents: ๋‚ด์šฉ(์ฃผ๋กœ post)๊ณผ ๊ด€๋ จ๋œ branch
  • designs: ์‚ฌ์ดํŠธ ๋””์ž์ธ ๋ฐ ๊ธฐ๋Šฅ(plugin)๊ณผ ๊ด€๋ จ๋œ branch

์—…๋ฐ์ดํŠธ

  • ์—…๋ฐ์ดํŠธ ํ•˜๋ ค๋Š” ๋‚ด์šฉ์— ๋”ฐ๋ผ, contents ๋˜๋Š” designs ์—์„œ branch๋ฅผ ์ƒˆ๋กœ ์ƒ์„ฑํ•˜์—ฌ commit ํ›„ ๋‹ค์‹œ PRํ•ฉ๋‹ˆ๋‹ค
  • PR ์™„๋ฃŒ ํ›„, release ๊ฒฐ์ •์— ๋”ฐ๋ผ contents, designs์„ master๋กœ mergeํ•ฉ๋‹ˆ๋‹ค.

ํฌ์ŠคํŠธ ์ž‘์„ฑ๋ฒ•

/blog ๋‚ด ๋‹ค๋ฅธ ํฌ์ŠคํŠธ๋“ค์„ ์ฐธ๊ณ ํ•˜์‹œ๋Š” ๊ฒŒ ๋น ๋ฆ…๋‹ˆ๋‹ค! ์ฐธ๊ณ : Docusaurus docs

1. ํฌ์ŠคํŠธ ์ƒ์„ฑ

  • blog/ ํด๋” ํ•˜์œ„์— Post๋ฅผ ๋‹ด์„ ํด๋”๋ช…์„ ์ƒ์„ฑํ•ฉ๋‹ˆ๋‹ค. ํ•ด๋‹น ํด๋”๋ช…์€ ๋‹ค๋ฅธ ํฌ์ŠคํŠธ์™€ ์–‘์‹๋งŒ ์–ผ์ถ” ๋น„์Šทํ•˜๊ฒŒ ๋งž์ถ”์‹œ๋ฉด ๋ฉ๋‹ˆ๋‹ค.
  • ํ•ด๋‹น ํด๋” ์•ˆ์— index.mdx (๋˜๋Š” index.md) ํŒŒ์ผ์„ ๋งŒ๋“ค๋ฉด, ํ•ด๋‹น ํŒŒ์ผ์„ ๊ธฐ์ค€์œผ๋กœ ํฌ์ŠคํŠธ๊ฐ€ ์ƒ์„ฑ๋ฉ๋‹ˆ๋‹ค.

::: note

index.mdx์™€ index.md์˜ ์ฐจ์ด๋Š” mdx ๋‚ด์— JSX๋กœ buildํ•˜๋Š” ๊ตฌ๋ฌธ์ด ์žˆ๋Š”์ง€ (์‰ฝ๊ฒŒ ์ด์•ผ๊ธฐํ•˜๋ฉด javascript ์ฝ”๋“œ๊ฐ€ ์žˆ๋Š” ์ง€)์— ๋”ฐ๋ฅธ ๊ฒƒ์œผ๋กœ ๊ฒฐ์ •๋ฉ๋‹ˆ๋‹ค. mdx ๊ฐ€ md์˜ ์ƒ์œ„ํ˜ธํ™˜์ธ ๋งŒํผ, ์ตœ์ข…์ ์œผ๋กœ push ํ•  ๋•Œ๋Š” mdx๋กœ ์ €์žฅํ•ด์ฃผ์‹œ๋ฉด ์ข‹์Šต๋‹ˆ๋‹ค.

:::

2. authors.yml ์— ๊ธฐ์ž…

Docusaurus๋Š” ์ €์ž๋ฅผ yaml ํŒŒ์ผ๋กœ ๊ด€๋ฆฌํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ๋‹ค๋ฅธ ๋ถ„๋“ค์˜ yaml ๋ณด๊ณ  ์ถ”๊ฐ€ํ•ด์ฃผ์„ธ์š”!

3. Front matter

.md ์ตœ์ƒ๋‹จ์— ์•„๋ž˜ ์˜ˆ์‹œ๋ฅผ ๋”ฐ๋ผ front matter๋ฅผ ๊ธฐ์ž…ํ•ฉ๋‹ˆ๋‹ค.

---
slug: nu-wave
title: "NU-Wave(Interspeech):"
description: ์ตœ์ดˆ๋กœ 48kHz๋กœ upsampling์„ ์„ฑ๊ณตํ•œ ์ €ํฌ ์—ฐ๊ตฌ๋ฅผ ์†Œ๊ฐœํ•ฉ๋‹ˆ๋‹ค.
image: img/maumai_Symbol.png
authors: [junjun3518, seungu]
tags: [publication, paper-review]
---
๋ณธ๋ฌธ...
  • slug: ์ฃผ์†Œ ์ฐฝ์— slash(/) ๋’ค์— ์–ด๋–ค ์ œ๋ชฉ์œผ๋กœ ๋ถ™์„ ์ง€๋ฅผ ๊ฒฐ์ •ํ•ฉ๋‹ˆ๋‹ค.
  • title: ํฌ์ŠคํŠธ ์ œ๋ชฉ
  • description: ํฌ์ŠคํŠธ์— ๋Œ€ํ•œ ์„ค๋ช…. ํฌ์ŠคํŠธ ์ž์ฒด์—์„œ๋Š” ํ‘œ์‹œ ๋˜์ง€ ์•Š์œผ๋‚˜, header์— ๋“ค์–ด๊ฐ€์„œ ์Šฌ๋ž™, ์นดํ†ก ๋“ฑ์— ๋ถ™์—ฌ ๋„ฃ์„ ๋•Œ preview๋กœ์„œ ํ‘œ์‹œ๋ฉ๋‹ˆ๋‹ค.
  • image: ํฌ์ŠคํŠธ์— ๋Œ€ํ•œ preview ๊ทธ๋ฆผ. ํฌ์ŠคํŠธ ์ž์ฒด์—์„œ๋Š” ํ‘œ์‹œ ๋˜์ง€ ์•Š์œผ๋‚˜, header์— ๋“ค์–ด๊ฐ€์„œ ์Šฌ๋ž™, ์นดํ†ก ๋“ฑ์— ๋ถ™์—ฌ ๋„ฃ์„ ๋•Œ preview๋กœ์„œ ํ‘œ์‹œ๋ฉ๋‹ˆ๋‹ค. ๋ชจ๋ฅด๊ฒ ์œผ๋ฉด img/maumai_Symbol.png ๋กœ ๋„ฃ์–ด์ฃผ์„ธ์š”.
  • tags: ๊ธ€์˜ ์นดํ…Œ๊ณ ๋ฆฌ๋ฅผ ์ž…๋ ฅํ•ฉ๋‹ˆ๋‹ค. ๋ชฉ๋ก ์ •ํ•ด์ง€๊ณ  ํ™•์žฅ๋  ์˜ˆ์ •.
    • publication
    • paper-review
    • news
    • etc ...
  • authors: ์ž‘์„ฑ์ž ์ด๋ฆ„์œผ๋กœ ํ‘œ์‹œ๋˜๋Š” ์ด๋ฆ„. authors.yml์˜ ํ‚ค ๊ฐ’์œผ๋กœ ์ž…๋ ฅํ•ด์ฃผ์‹œ๋ฉด ๋ฉ๋‹ˆ๋‹ค.

3. ๋ณธ๋ฌธ ์ž‘์„ฑ

  • markdown ํ˜•์‹์„ ๋”ฐ๋ผ ์ž‘์„ฑํ•ฉ๋‹ˆ๋‹ค.
  • ์ œ๋ชฉ์€ front matter์˜ title์ด ์ž๋™์œผ๋กœ ๋ Œ๋”๋ง๋˜๊ณ , ์ดํ›„ ๋ถ€์ œ๋ชฉ์€ ##๋กœ, ์†Œ์ œ๋ชฉ์€ ###๋กœ ์ž…๋ ฅํ•ฉ๋‹ˆ๋‹ค.
  • ํฌ์ŠคํŠธ ์ค‘์— <!--truncate--> ๋ฅผ ์ž…๋ ฅํ•˜๋ฉด, ํ•ด๋‹น ๊ธ€์„ preview ํ•  ๋•Œ <!--truncate--> ์ง์ „๊นŒ์ง€์˜ ๋ถ€๋ถ„๋งŒ ๋ณด์ž…๋‹ˆ๋‹ค. ์ฒซ ๋ฌธ๋‹จ์— Contribution์„ ์ž‘์„ฑํ•˜๊ฑฐ๋‚˜ ์ธ์‚ฌ๋ฅผ ๋‚จ๊ธฐ์‹œ๊ณ , ๋‹ค์Œ ๋ฌธ๋‹จ ์˜ค๊ธฐ ์ „์— ๋„ฃ์–ด์ฃผ์‹œ๋ฉด ์ œ์ผ ์ข‹์„ ๊ฒƒ ๊ฐ™์•„์š”!

4. ์ด๋ฏธ์ง€ ๋„ฃ๊ธฐ

  • ์ž๊ธฐ ํด๋” ๋‚ด์— image ํด๋”๋กœ ํŒŒ์ผ์„ ์œ„์น˜์‹œํ‚ต๋‹ˆ๋‹ค.
  • ํ˜„์žฌ ์ด๋ฏธ์ง€๋ฅผ javascript๋กœ ์ผ์ผํžˆ ๋ Œ๋”๋งํ•˜์—ฌ center align ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋‹ค๋ฅธ ํฌ์ŠคํŠธ์—์„œ ์ด๋ฏธ์ง€๋ฅผ ๋„ฃ๋Š” ๋ฐฉ๋ฒ•์„ ์ฐธ๊ณ ํ•˜์‹œ์–ด ์ง„ํ–‰ํ•ด์ฃผ์‹œ๋ฉด ๊ฐ์‚ฌํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค.
  • ๋ชจ๋ฅด๊ฒ ๋‹ค ์‹ถ์œผ์‹ค ๋•Œ๋Š” ๊ทธ๋ƒฅ alt text ๋กœ ๋„ฃ์–ด์ฃผ์‹œ๊ณ  contents ๋กœ push ํ•ด์ฃผ์‹œ๋ฉด Tech Blog ํŒ€์ด ์•Œ์•„์„œ ํ•ด์ค„ ๊ฒ๋‹ˆ๋‹ค.

5. References ์ž‘์„ฑ

  • Docusaurus Rendering ์ดํ›„ ๊ธฐ์กด์˜ ์†์‰ฌ์šด reference ๋ฐฉ์‹์ด ์ž˜ ๋˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค. ๋‹ค๋ฅธ ๊ธ€๋“ค์„ ์ฐธ๊ณ ํ•˜์—ฌ ์ง„ํ–‰ํ•ด์ฃผ์‹œ๋ฉด ๊ฐ์‚ฌํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค.

6. Badge ๋‹ฌ๊ธฐ

  • ์ฐธ๊ณ : shields.io
  • (๋˜๋„๋ก ํฌ์ŠคํŠธ ์œ—๋ถ€๋ถ„์—์„œ) ํฌ์ŠคํŠธ์—์„œ ์ฃผ๋กœ ๋‹ค๋ฃจ๋Š” ๋Œ€์ƒ์˜ ๋งํฌ๋“ค์„ ์•„๋ž˜์™€ ๊ฐ™์ด badge๋กœ ๋‹ฌ์•„๋‘๋ฉด ๋ณด๊ธฐ์— ์ข‹์Šต๋‹ˆ๋‹ค.
  • arXiv CVF GitHub Repo stars githubio githubio Colab
    ## Awesome NU-WAVE!
    
    ### It has many public links
    ...
    [![arXiv](https://img.shields.io/badge/arXiv-2104.02321-brightgreen.svg?style=flat-square)](https://arxiv.org/abs/2104.02321)
    [![CVF](https://img.shields.io/badge/CVF-2021.15059-9cf.svg?style=flat-square)](https://openaccess.thecvf.com/content/CVPR2021/html/Kim_SetVAE_Learning_Hierarchical_Composition_for_Generative_Modeling_of_Set-Structured_Data_CVPR_2021_paper.html)
    [![GitHub Repo stars](https://img.shields.io/github/stars/mindslab-ai/nuwave?color=yellow&label=nu-wave&logo=github&style=flat-square)](https://github.com/mindslab-ai/nuwave)
    [![githubio](https://img.shields.io/badge/GitHub.io-audio_samples-blue?logo=Github&style=flat-square)](https://mindslab-ai.github.io/nuwave/)
    [![githubio](https://img.shields.io/static/v1?message=Official%20Repo&logo=Github&labelColor=grey&color=blue&logoColor=white&label=%20&style=flat-square)](https://github.com/mindslab-ai/nuwave)
    [![Colab](https://img.shields.io/static/v1?message=Open%20in%20Colab&logo=googlecolab&labelColor=grey&color=yellow&logoColor=white&label=%20&style=flat-square)](https://colab.research.google.com/drive/1AK3AI3lS_rXacTIYHpf0mYV4NdU56Hn6?usp=sharing)
    
    ### The author is awesome
    ...
    

7. ์ˆ˜์‹ ์ž…๋ ฅ

๋ฐฉ๋ฒ• 1: block equation

$$
\operatorname{swish}(x):=x \times \sigma(\beta x)=\frac{x}{1+e^{-\beta x}}
$$

๋ฐฉ๋ฒ• 2: inline

  • $\beta = 0$ ์ผ ๊ฒฝ์šฐ, Linear function $f(x) = x/2$ ์ฒ˜๋Ÿผ ์ž‘์šฉํ•˜๊ฒŒ ๋ฉ๋‹ˆ๋‹ค.inline์œผ๋กœ ์ž‘์—…ํ• ๋•Œ๋Š” ${something}$ ๊ณผ ๊ฐ™์€ ์‹์œผ๋กœ ์ž‘์„ฑ๊ฐ€๋Šฅ

8. Emoji ์ž…๋ ฅ

  • emoji๋ฅผ ๋ณต์‚ฌํ•ด์„œ ๋ถ™์—ฌ ๋„ฃ์œผ๋ฉด ๋“ค์–ด๊ฐ€๋‚˜ (์œ ๋‹ˆ์ฝ”๋“œ ๋ฐฉ์‹), ์Šฌ๋ž™์ฒ˜๋Ÿผ :{emoji}: ํ˜•์‹์€ ์ง€์›๋˜์ง€ ์•Š์Šต๋‹ˆ๋‹ค.
  • ์ฐธ๊ณ : https://apps.timwhitlock.info/emoji/tables/unicode

Publication ์ถ”๊ฐ€ํ•˜๊ธฐ

BrainํŒ€์— ๋˜ ํ•˜๋‚˜์˜ ๋…ผ๋ฌธ์ด ์ƒ๊ฒผ๊ตฐ์š”! ์•„๋ž˜ ๋‚ด์šฉ์„ ํ™•์ธํ•˜์…”์„œ Tech Blog์— ์ž๋ž‘์Šค๋Ÿฌ์šด ๋…ผ๋ฌธ์„ ์ถ”๊ฐ€ํ•ด์ฃผ์„ธ์š”๐Ÿ˜€

0. ์˜ˆ์‹œ ํ™•์ธํ•˜๊ธฐ

์•„๋ž˜ ๋˜๋Š” publications.mdx์—์„œ Publication ๋„ฃ๋Š” ์˜ˆ์‹œ๋ฅผ ํ™•์ธํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

<li>
  <features.ConferenceItem conference="Interspeech"/>
  <features.PaperTitle paperLink="https://arxiv.org/abs/2206.08545" title="NU-Wave 2: A General Neural Audio Upsampling Model for Various Sampling Rates"/>
  <features.AuthorItem authors={["Seungu Han", "Junhyeok Lee"]} numFirstAuthor={1} isBrainTeam={[true, true]}/>
  <features.PaperDescription preview="Conventionally, audio super-resolution models fixed the initial and the target sampling rates, which necessitate the model to be trained for each pair of sampling rates. "
  description="We introduce NU-Wave 2, a diffusion model for neural audio upsampling that enables the generation of 48 kHz audio signals from inputs of various sampling rates with a single model. Based on the architecture of NU-Wave, NU-Wave 2 uses short-time Fourier convolution (STFC) to generate harmonics to resolve the main failure modes of NU-Wave, and incorporates bandwidth spectral feature transform (BSFT) to condition the bandwidths of inputs in the frequency domain. We experimentally demonstrate that NU-Wave 2 produces high-resolution audio regardless of the sampling rate of input while requiring fewer parameters than other models."/>
  <features.GithubItem link="https://github.com/mindslab-ai/nuwave2" />
  <features.DemoItem link="https://mindslab-ai.github.io/nuwave2/" />
</li>

1. ํ•™ํšŒ ๋„ฃ๊ธฐ

ํ•™ํšŒ๋ฅผ ์•„๋ž˜์™€ ๊ฐ™์ด ์ž…๋ ฅํ•ด์ฃผ์‹œ๋ฉด ๋ฉ๋‹ˆ๋‹ค.

<features.ConferenceItem conference="Interspeech"/>

Attribute ๋ชฉ๋ก

  • conference: ํ•™ํšŒ ์ด๋ฆ„์„ ์ž…๋ ฅํ•ฉ๋‹ˆ๋‹ค. ํ•™ํšŒ๋…„๋„๋Š” ์ƒ๋žตํ•ด์ฃผ์„ธ์š”. Oral Paper๋กœ ์„ ์ •๋˜๋Š” ๋“ฑ ์ž๋ž‘์Šค๋Ÿฌ์šด ์„ฑ๊ณผ๊ฐ€ ์žˆ์œผ์‹œ๋ฉด, (Oral)์„ ์ถ”๊ฐ€ํ•ด์ฃผ์…”๋„ ๋ฉ๋‹ˆ๋‹ค.

2. ์ œ๋ชฉ ๋„ฃ๊ธฐ

์ œ๋ชฉ์„ ์•„๋ž˜์™€ ๊ฐ™์ด ์ž…๋ ฅํ•ด์ฃผ์‹œ๋ฉด ๋ฉ๋‹ˆ๋‹ค.

<features.PaperTitle paperLink="https://arxiv.org/abs/2206.08545" title="NU-Wave 2: A General Neural Audio Upsampling Model for Various Sampling Rates"/>

Attribute ๋ชฉ๋ก

  • paperLink: ๋…ผ๋ฌธ ๋งํฌ๋ฅผ ๋„ฃ์–ด์ฃผ์‹œ๋ฉด ๋ฉ๋‹ˆ๋‹ค. arXiv ๋งํฌ๋„ ๊ฐ€๋Šฅํ•ฉ๋‹ˆ๋‹ค.
  • title: ๋…ผ๋ฌธ ์ œ๋ชฉ์„ ์ž…๋ ฅํ•ด์ฃผ์„ธ์š”.

3. ์ €์ž ๋„ฃ๊ธฐ

์ €์ž๋ฅผ ์•„๋ž˜์™€ ๊ฐ™์ด ์ž…๋ ฅํ•ด์ฃผ์‹œ๋ฉด ๋ฉ๋‹ˆ๋‹ค.

<features.AuthorItem authors={["Seungu Han", "Junhyeok Lee"]} numFirstAuthor={1} isBrainTeam={[true, true]}/>

AuthorItem ์€ ์•„๋ž˜๋ฅผ ์ž๋™์œผ๋กœ ์ฒ˜๋ฆฌํ•ด์ค๋‹ˆ๋‹ค.

  • ๊ณต๋™ 1์ €์ž ์ˆ˜๋งŒํผ * ํ‘œ์‹œ๋ฅผ ํ†ตํ•ด 1์ €์ž๋ฅผ ํ‘œ์‹œํ•ด์ค๋‹ˆ๋‹ค. (numFirstAuthor)
  • MINDsLab BrainํŒ€ ์—ฌ๋ถ€์— ๋”ฐ๋ผ ๋ณผ๋“œ ํ‘œ์‹œ๋ฅผ ํ•  ์ง€ ๊ฒฐ์ •ํ•ฉ๋‹ˆ๋‹ค. (isBrainTeam)
  • ๊ณต์ €์ž๊ฐ€ 1๋ช…, 2๋ช…, 3๋ช… ์ด์ƒ์ธ ์ƒํ™ฉ์— ๋”ฐ๋ผ and, , ๋“ฑ์˜ ์ถ”๊ฐ€๋ฅผ Oxford Comma๋ฅผ ๋”ฐ๋ผ ์ž๋™์œผ๋กœ ํ•ด์ค๋‹ˆ๋‹ค.

Attribute ๋ชฉ๋ก

  • authors: ์ €์ž ๋ชฉ๋ก List๋ฅผ ํ•„๋ช…์œผ๋กœ ๋„ฃ์–ด์ฃผ์‹œ๋ฉด ๋ฉ๋‹ˆ๋‹ค. Array ์ธ์‹์„ ์œ„ํ•ด {}๋กœ ๊ฐ์‹ธ์ฃผ์„ธ์š”.
  • numFirstAuthor: ๊ณต๋™ 1์ €์ž ์ˆ˜๋ฅผ ์ž…๋ ฅํ•ด์ฃผ์„ธ์š”. ์ €์ž ๋ชฉ๋ก ๋งจ ์•ž๋ถ€ํ„ฐ ํ•ด๋‹น ์ˆ˜ ๋งŒํผ 1์ €์ž ํ‘œ๊ธฐ๋ฅผ ํ•ฉ๋‹ˆ๋‹ค. Integer ์ธ์‹์„ ์œ„ํ•ด {}๋กœ ๊ฐ์‹ธ์ฃผ์„ธ์š”.
  • isBrainTeam: ๊ฐ ์ €์ž ๋ถ„์ด MINDsLab BrainํŒ€์ด์‹ ์ง€ ํ‘œ๊ธฐํ•ด์ฃผ์„ธ์š”. ๊ผญ ์ €์ž ์ธ์› ์ˆ˜์™€ ๋™์ผํ•˜๊ฒŒ Array์— boolean ๊ฐ’์„ ๋„ฃ์–ด์ฃผ์„ธ์š”. Array ์ธ์‹์„ ์œ„ํ•ด {}๋กœ ๊ฐ์‹ธ์ฃผ์„ธ์š”.

4. Abstract ๋„ฃ๊ธฐ

์•„๋ž˜์™€ ๊ฐ™์ด ๋„ฃ์–ด์ค๋‹ˆ๋‹ค.

<features.PaperDescription preview="Conventionally, audio super-resolution models fixed the initial and the target sampling rates, which necessitate the model to be trained for each pair of sampling rates. "
description="We introduce NU-Wave 2, a diffusion model for neural audio upsampling that enables the generation of 48 kHz audio signals from inputs of various sampling rates with a single model. Based on the architecture of NU-Wave, NU-Wave 2 uses short-time Fourier convolution (STFC) to generate harmonics to resolve the main failure modes of NU-Wave, and incorporates bandwidth spectral feature transform (BSFT) to condition the bandwidths of inputs in the frequency domain. We experimentally demonstrate that NU-Wave 2 produces high-resolution audio regardless of the sampling rate of input while requiring fewer parameters than other models."/>

Attribute ๋ชฉ๋ก

  • preview: Show More ๋ˆ„๋ฅด๊ธฐ ์ด์ „์— ๋ณด์—ฌ์ง€๋Š” ๋‚ด์šฉ์„ ์ž…๋ ฅํ•ด์ฃผ์„ธ์š”. ์ผ๋ฐ˜์ ์œผ๋กœ Abstarct ์ฒซ ๋ฌธ์žฅ๋งŒ ์ž…๋ ฅํ•˜๋Š” ๊ฒƒ์„ ์ถ”์ฒœํ•ฉ๋‹ˆ๋‹ค.
  • description: preview์— ๋„ฃ์€ ๋ถ€๋ถ„์„ ์ œ์™ธํ•œ ๋‚˜๋จธ์ง€ Abstract๋ฅผ ์ž…๋ ฅํ•ด์ฃผ์„ธ์š”. ํŽ˜์ด์ง€ ๋ฐฉ๋ฌธ์ž๊ฐ€ Show More์„ ๋ˆŒ๋Ÿฌ์•ผ ๋ณด์ž…๋‹ˆ๋‹ค.

5. Supplements ๋„ฃ๊ธฐ

Github link, demo link ๋“ฑ์„ ์•„๋ž˜์™€ ๊ฐ™์ด ๋„ฃ์–ด์ค๋‹ˆ๋‹ค. Code๋‚˜ demo๊ฐ€ ์•„๋‹Œ ๊ฒฝ์šฐ (ex. Screencast), MiscItem์„ ์ด์šฉํ•˜์—ฌ ์ถ”๊ฐ€ํ—ค์ฃผ์„ธ์š”.

<features.GithubItem link="https://github.com/mindslab-ai/nuwave2" />
<features.DemoItem link="https://mindslab-ai.github.io/nuwave2/" />
<features.DemoItem link="https://huggingface.co/spaces/CVPR/ml-talking-face" customName="๐Ÿค—Demo" />
<features.MiscItem link="https://www.youtube.com/watch?v=toqdD1F_ZsU" customName="Screencast" />

CSS ๋‚ด์—์„œ ์Šคํƒ€์ผ์€ GithubItem, DemoItem, MiscItem์— ๋”ฐ๋ผ ๋‹ค๋ฅด๊ฒŒ ์ฒ˜๋ฆฌ๋  ์ˆ˜ ์žˆ๊ธฐ๋Š” ํ•˜๋‚˜, ํ˜„์žฌ๋Š” ์Šคํƒ€์ผ์ด ๋™์ผํ•˜๊ฒŒ ์ ์šฉ๋˜์–ด ์žˆ์Šต๋‹ˆ๋‹ค.

Attribute ๋ชฉ๋ก

GithubItem, DemoItem, MiscItem ๋™์ผํ•œ attribute๋ฅผ ๊ฐ€์ง€๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.

  • link: ๋ถ€๊ฐ€ ์ž๋ฃŒ๋กœ ๊ฐ€๋Š” ๋งํฌ๋ฅผ ์ž…๋ ฅํ•ฉ๋‹ˆ๋‹ค.
  • customName (์„ ํƒ, ํ•„์ˆ˜): ํŽ˜์ด์ง€์—์„œ ํ‘œ์‹œ๋˜๋Š” ์ด๋ฆ„์„ ๋ณ€๊ฒฝํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. MiscItem์—์„œ๋Š” ํ•„์ˆ˜๋กœ ์ž…๋ ฅํ•ด์•ผ ํ•˜๋ฉฐ, ๊ทธ ์™ธ์—์„œ๋Š” ์„ ํƒ์ž…๋‹ˆ๋‹ค. ์ž…๋ ฅํ•˜์ง€ ์•Š์„ ๊ฒฝ์šฐ, GithubItem์€ Github, DemoItem์€ Demo๋กœ ํ‘œ์‹œ๋ฉ๋‹ˆ๋‹ค.

Open Source ์ถ”๊ฐ€ํ•˜๊ธฐ

Open source์˜ ๊ฒฝ์šฐ, MINDsLab ๊ณต์‹ Github์— ์ถ”๊ฐ€ํ•˜๋Š” ๊ฒƒ์„ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค. ๋…ผ๋ฌธ ๋ฐœํ‘œ์™€ ํ•จ๊ป˜ ์ฝ”๋“œ ๊ณต๊ฐœ๋ฅผ ํ•˜๋Š” ๊ฒฝ์šฐ์—๋Š” Official Repo๋กœ, ๋‹ค๋ฅธ ๋…ผ๋ฌธ์„ ๋ณด๊ณ  ๊ตฌํ˜„ํ•œ ์ฝ”๋“œ์˜ ๊ฒฝ์šฐ Unofficial Repo๋กœ ๋“ฑ๋กํ•ด์ฃผ์„ธ์š”.

0. ์˜ˆ์‹œ ํ™•์ธํ•˜๊ธฐ

์•„๋ž˜ ๋˜๋Š” open-source.mdx์—์„œ open source ๋„ฃ๋Š” ์˜ˆ์‹œ๋ฅผ ํ™•์ธํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค.

<li>
  <features.StarItem repoName="pnlp-mixer" />
  <features.GithubLinkItem repoName="pnlp-mixer" repoNickname="pNLP-Mixer"  />
  <features.PaperLinkItem paperLink="https://arxiv.org/abs/2202.04350" title="pNLP-Mixer: an Efficient all-MLP Architecture for Language" />
  <p className={styles.description}>
      First successful open-source implementation of <i>pNLP-Mixer</i>.
  </p>
</li>

1. Github ๋„ฃ๊ธฐ

Github repo๋ฅผ ๋„ฃ๊ธฐ ์œ„ํ•ด์„œ๋Š” StarItem๊ณผ GithubLinkItem์„ ๊ฐ๊ฐ ์ˆœ์„œ๋Œ€๋กœ ์ž…๋ ฅํ•ด์ฃผ์…”์•ผ ํ•ฉ๋‹ˆ๋‹ค.
๊ถŒ์žฅํ•˜์ง€๋Š” ์•Š์œผ๋‚˜, ๋งŒ์•ฝ open source๊ฐ€ Github๊ฐ€ ์•„๋‹Œ ๊ณณ (ex. Bitbucket)์— push ๋˜์–ด ์žˆ์„ ๊ฒฝ์šฐ, ์ˆ˜๋™์œผ๋กœ ์ž…๋ ฅํ•ด์ฃผ์…”์•ผ ํ•ฉ๋‹ˆ๋‹ค.

/* MINDsLab ๊ณต์‹ Github์— ์žˆ๋Š” Repo ์ถ”๊ฐ€ํ•  ๊ฒฝ์šฐ */
<features.StarItem repoName="pnlp-mixer" />
<features.GithubLinkItem repoName="pnlp-mixer" repoNickname="pNLP-Mixer"  />

/* ๊ฐœ์ธ Github์— ์žˆ๋Š” Repo ์ถ”๊ฐ€ํ•  ๊ฒฝ์šฐ */
<features.StarItem userName="seungwonpark" repoName="melgan" />
<features.GithubLinkItem userName="seungwonpark" repoName="melgan" repoNickname="MelGAN" />

Attribute ๋ชฉ๋ก

StarItem๊ณผ GithubLinkItem์€ ์•„๋ž˜ attribute๋ฅผ ๋™์ผํ•˜๊ฒŒ ๊ฐ–๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.

  • userName (์„ ํƒ): Repo ์†Œ์œ ์ž์˜ Github username์„ ์ž…๋ ฅํ•ฉ๋‹ˆ๋‹ค. ์ž…๋ ฅํ•˜์ง€ ์•Š์„ ๊ฒฝ์šฐ, mindslab-ai๋กœ ์ž…๋ ฅ๋˜์–ด ์ž๋™์ ์œผ๋กœ ๊ณต์‹ Github์— ์žˆ๋Š” repo๋ฅผ ๊ฐ€์ ธ์˜ต๋‹ˆ๋‹ค.
  • repoName: Repo ์ด๋ฆ„์„ ์ž…๋ ฅํ•ฉ๋‹ˆ๋‹ค.

GithubLinkItem์€ ์•„๋ž˜ attribute๋ฅผ ์ถ”๊ฐ€์ ์œผ๋กœ ๊ฐ€์ง€๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.

  • repoNickname: ํŽ˜์ด์ง€์— ํ‘œ์‹œ๋  Repo ์ด๋ฆ„์„ ์ž…๋ ฅํ•ด์ฃผ์„ธ์š”.

2. ๋…ผ๋ฌธ ๋งํฌ ๋„ฃ๊ธฐ

๋…ผ๋ฌธ ๊ตฌํ˜„์ฒด์˜ ๊ฒฝ์šฐ, ๋…ผ๋ฌธ ๋งํฌ๋ฅผ ํ•จ๊ป˜ ํ‘œ์‹œํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์•„๋ž˜์™€ ๊ฐ™์ด ๋„ฃ์–ด์ฃผ์„ธ์š”.

<features.PaperLinkItem paperLink="https://arxiv.org/abs/2206.08545" title="NU-Wave 2: A General Neural Audio Upsampling Model for Various Sampling Rates"/>

Attribute ๋ชฉ๋ก

  • paperLink: ๋…ผ๋ฌธ ๋งํฌ๋ฅผ ๋„ฃ์–ด์ฃผ์‹œ๋ฉด ๋ฉ๋‹ˆ๋‹ค. arXiv ๋งํฌ๋„ ๊ฐ€๋Šฅํ•ฉ๋‹ˆ๋‹ค.
  • title: ๋…ผ๋ฌธ ์ œ๋ชฉ์„ ์ž…๋ ฅํ•ด์ฃผ์„ธ์š”.

3. ์„ค๋ช… ๋„ฃ๊ธฐ

Official์˜ ๊ฒฝ์šฐ, ๋ณ„๋„์˜ ์„ค๋ช…์ด ํ•„์š”ํ•˜์ง€๋Š” ์•Š์œผ๋‚˜, Unofficial์˜ ๊ฒฝ์šฐ, ์ถ”๊ฐ€ ์„ค๋ช…์„ ๋„ฃ๋Š” ๊ฒƒ์„ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค. Github repo ๋‚ด About๊ณผ ๋™์ผํ•˜๊ฒŒ ์ž‘์„ฑํ•˜๋Š” ๊ฒƒ์„ ๊ถŒ์žฅํ•ฉ๋‹ˆ๋‹ค.
๋ณธ ์‚ฌํ•ญ์€ HTML๋กœ ์ž์œ ๋กญ๊ฒŒ ์ž…๋ ฅํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ์•„, ๋ณ„๋„๋กœ ํ•จ์ˆ˜ ์ž‘์—…์„ ์ง„ํ–‰ํ•˜์ง€ ์•Š์•˜์Šต๋‹ˆ๋‹ค. className๋งŒ ๋™์ผํ•˜๊ฒŒ ๋งž์ถ”์‹œ๊ณ , ๋‚ด์šฉ์€ ์ž์œ ๋กญ๊ฒŒ ์ž…๋ ฅํ•ด์ฃผ์‹œ๋ฉด ๊ฐ์‚ฌํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค.

<p className={styles.description}>
    First successful open-source implementation of <i>pNLP-Mixer</i>.
</p>